fields Spatiotemporal organization of receptive ... - Semantic Scholar
Quantitative studies of single-cell properties in monkey striate cortex. I. Spatiotemporal organization of receptive fields P. H. Schiller, B. L. Finlay and S. F. Volman J Neurophysiol 39:1288-1319, 1976. ;
This article has been cited by 56 other HighWire-hosted articles: http://jn.physiology.org/content/39/6/1288#cited-by Updated information and services including high resolution figures, can be found at: http://jn.physiology.org/content/39/6/1288.full Additional material and information about Journal of Neurophysiology can be found at: http://www.the-aps.org/publications/jn This information is current as of June 3, 2013.
Journal of Neurophysiology publishes original articles on the function of the nervous system. It is published 12 times a year (monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 1976 the American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at http://www.the-aps.org/.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
You might find this additional info useful...
JOURNAL. OFNEUROPHYSIOLOGY Vol. 39, No. 6, November 1976. Printed
in U.S.A.
PETER
H. SCHILL,ER,
Depurtment Cambridge,
SUMMARY
BARBARA
of Psychology, Massachusetts
AND
L,. FINLAY,
Massachusetts 02139
Institute
CONCLUSIONS
1. The properties of single cells in striate cortex of the rhesus monkey, representing the visual field 2”-5” from the fovea, were examined quantitatively with stationary and moving stimuli. Three distinct classes of cells were identified: S type, CX type, and T type. 2. S-type cells were defined as those oriented cells which to the optimal direction of movement in their receptive fields exhibited one or more spatially separate subfields within each of which a response was obtained to either a light or dark edge, but not to both. Several different types of S-cells were distinguished: u) S-type cells for which moving edges revealed a single excitatory area within which a response was elicited by either a light or a dark edge but not by both. Most of these cells were unidirectional. 6) S,type cells for which moving edges revealed two spatially separate response areas, one of which was excited by a light edge and the other by a dark edge. Both regions responded to the same direction of movement. c) &-type cells which had two response areas, one of which was excited by a stimulus moving in one direction (at right angles to the axis of orientation) and the other, of opposite contrast, which responded in the opposite direction. (i) S-type cells which to one direction of movement showed two spatially separate regions sensitive to a light and dark edge and which in the other direction of movement had only one responsive area (either light or dark). c) Cells which had multiple spatially separate subfields (S,., types). 3. CX-type cells were defined as those oriented cells which in their receptive fields exhibited no spatial separation for light- and dark-edge responses; they discharged to both edges in the same direction of movement and in Received 1288
for
publication
December
22, 1975.
AND
SUSAN
F. VOLMAN
of Technology,
the same spatial area. Flashing stimuli elicited both on and off responses throughout the receptive field. CX-type cells were predominantly of two types: those which were selective for direction of stimulus movement and those which were not. 4. A third class of cells (T-type) were those which were excited by only one sign of contrast change and responded in a sustained fashion even when there was no contour within the receptive field. These cells were poorly or not at all oriented; some of them were selective to wavelength. 5. Quantitative comparisons showed the following differences between S-type and CX-type cells: u) S-type cells had smaller receptive fields than CX-type cells but the populations overlapped considerably. Receptive-field size was smallest in layer 4c. In all other layers S-type cells had the same size fields. CX-type cells, by contrast, tended to have larger fields in layer S-6 than 2-3. b) The spatial separation between light and dark response areas was the best criterion for distinguishing S-type and CX-type cells. The distribution of this measure disclosed two populations of cells with relatively limited overlap. c) In layers 2 and 3, both S-type and CX-type cells had low spontaneous activity. In layers 5 and 6, S-type cells retained this attribute. Most CXtype cells, however, showed significantly higher spontaneous activity in these bottom layers. Thus the criterion of spontaneous activity discriminated S-type and CX-type cells in layers 5 and 6 but failed to do so in the upper layers. cl) S-type cells were found in all cortical layers. CX-type cells were absent in layer 4c. 6. The measure of inhibition along the axis of orientation, typically used to classify cells into the “hypercomplex” category, did not establish a distinct subcategory of cells. This property was observed in both S-type and CX-type cells
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
Quantitative Studies of Single-Cell Properties in Monkey Striate Cortex. I. Spatiotemporal Organization of Receptive Fields
MONKEY
STRIATE
INTRODUCTION
Following the pioneering work of Hubel and Wiesel (7, 8), striate cortex has been subjected to extensive analysis with single-unit recording techniques (l-5, 1 l-14). The input to this structure from the lateral geniculate nucleus (LGN) is transformed in five notable ways. The first of these produces specificity for the orientation of contours within the receptive field. The second transformation is that of direction selectivity. While in the LGN of the cat and monkey cells respond to stimuli moving in any direction, in cortex many cells are responsive to movement only in one direction. The third transformation involves the combination of responses to contrasts of opposite sign. Most LGN cells in the center of their receptive field are excited by either light increment or light decrement, but not by both. In cortex these attributes are combined so that many cells respond to both kinds of stimulus change within a single region. The fourth transformation unites the input from the two eyes, resulting in cells which can be binocularly activated. The fifth transformation alters the spatial frequency response of cells: in the striate cortex there is a sharply increased selectivity for spatial frequency when compared to what is found in the LGN (14). All five transformations may be seen in a single cell. In this series of papers we will present the results of experiments in which we analyzed the visual cortex of the monkey in terms of these transformations. In the first paper we will examine the response properties of single neurons in striate cortex *to direction of movement, contrast, and stimulus length, with special emphasis on simple cells. The second paper will present data regarding orientation selectivity and ocular dominance. The third paper will be concerned with selectivity for spatial frequency. The fourth paper will examine the properties of the corticotectal cells in area 17. In the last paper we will consider the relationship among the different measures we used and how these m easure s reflec t on various models of visual corte X. Work on the cat has disc losed several cri teria whe rcby simple and compl ex cells can be distinguis hed (1, 8, 20). Simple cells have relatively
I
1289
small receptive fields. These fields can be plotted with stationary flashing spots of light; such a plot discloses spatially separate activating regions for the onset (on-response) and the turning off (off-response) of a flashing light. The response within a given region summates as the size of the stimulus increases. Moving stimuli may demonstrate either one such activating region or several spatially separate activating regions. Furthermore, simple cells have low spontaneous activity and show a preference for relatively slow velocities of stimulus movement. In the cat, cells with simple receptive fields are most numerous in layers 3, 4, and 6 of striate cortex (8). Complex cells have larger fields and respond more poorly to flashes. The activating region shows no subdivision into spatially separate zones: when a cell responds to them, flashing stimuli elicit both on- and off-responses throughout the field; similarly, a response is obtained to both the light edge and the dark edge of moving stimuli within this one region. These cells are more often binocular, tend to have greater spontaneous activity, especially in cortical layers 5 and 6, and respond over a greater range of velocities. In contrast to work on the cat, simple cells in the monkey have not been studied in detail, primarily because they have been encountered less frequently. Hubel and Wiesel (10) report that only about 8% (25 of 272) of the cells they have studied in the monkey fell into this category. Poggio (15) classified 7% (12 of 224) and Dow (5) 9% (21 of 234) of the cells as simple. Considerable uncertainty exists regarding the location of these cells in monkey striate cortex. Hubel and Wiesel (10) found that of the 25 cells they classified as simple, 23 were in layer 4. By contrast, Poggio (15) reports only 1 of 12 in this layer. Dow (5) shows a relatively even distribution of simple cells among the cell layers. According to Hubel and Wiesel (IO), complex cells in the monkey are least common in layers 4b and 4c. Poggio’s data show them equally distributed. Subsequent to their initial discoveries, a third class of cells has been identified by Hubel and Wiesel (9) in the cat; these cells were called hypercomplex, and it was suggested that they were formed by excitatory and inhibitory inputs from complex cells. The major criterion for placing the hypercomplex cells into a distinct class was the observation that their responses are inhibited by stimuli extended along the axis of orientation, thereby rendering them selective for stimulus length, an attribute absent in simple and complex cells. Several qualifications have been made subsequently regarding this cell type (3). Further investigation disclosed several other
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
and was continuously distributed. However, this inhibition was more prevalent in the superficial layers than in the deep ones. 7. The results are suggestive of a hierarchy among S-type cells. Lowestin this hierarchy are those S-type cells which could be activated by only one kind of moving edge. From these, S-type cells with two or more subfields may be constructed.
CORTEX
1290
SCHILLER,
FINLAY,
VOLMAN
which within their activating region showed no spatial separation; the areas which responded to light increment and light decrement overlapped. Specificity for direction in these cells was similar for moving light edges (light increment) and dark edges (light decrement). In what follows we present qualitative and quantitative data describing the properties of various types of cells found in monkey striate cortex as determined by moving edges and stationary, flashing stimuli. We will place special emphasis on S-type cells as here defined. Distributions of cells according to a variety of response measures will be used to scrutinize the validity of classificatory schemes. Finally, these data will be considered in terms of the anatomical distribution of cell types within the layers of striate cortex. METHODS
Single-unit recording procedures A total of 42 rhesus monkeys (Macaca mulatta) were used to obtain the data reported in this paper. All 42 were studied while the animals were anesthetized and paralyzed. In addition, two of these animals had also been recorded from while alert, with one eye surgically immobilized. ANESTHETIZED ANIMALS. Surgical procedures were carried out under Pentothal anesthesia. After tracheotomy and vein cannulation, animals were placed in a Kopf stereotaxic apparatus with raised eye and ear bars. In 32 animals a closed chamber of 10 or 16 mm diameter was implanted over area 17 in a region where receptive fields were found 2”-5” from the fovea in the lower visual field. The dura mater overlying the visual cortex was removed in approximately half the animals and was left intact in the others. We found that successive penetrations can be undertaken with greater ease, rapidity, and with fewer episodes of electrode breakage when dura is removed. This procedure also permits the use of finer electrodes. On the other hand, recordings are not quite as stable and cortex deteriorates more rapidly. In 32 of the animals penetrations were made perpendicular to the surface of the brain. In 10 monkeys tangential penetrations were made which were between 10” and 25” to the surface of the brain, in a manner similar to that described by Hubel and Wiesel (12). Animals were infused with Flaxedil (40 mg/h) dissolved in 5% dextrose in Ringer and were artificially respired. End-tidal CO, was monitored and maintained at 4.&4.5%. Pentothal anesthesia was discontinued after the first 2-10 h
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
classes of cells in striate cortex. Prominent among these are unoriented cells, which may be further divisible into several subgroups (5, 10, 15). One of the problems with these classifications is that few of the dimensions which have been used for assigning cells to categories have been studied quantitatively. Thus it is not known, especially in the monkey, to what extent the measurements used form distributions which justify the classificatory schemes. The assumption of classes can be inferred quantitatively when the measure on the basis of which the classification is made yields separate groupings as demonstrated, for example, by a bimodal distribution. If, however, a distribution is continuous, one cannot, of course, claim two distinct classes. An example in point is binocularity. Since the extent to which cortical cells are binocular forms a relatively continuous distribution, this property has been believed not to reflect two distinct classes of monocular and binocular cells. Thus, issues about binocularity are typically formulated in statistical terms asking, for instance, how the ocular dominance distribution is affected by such variables as brain site or environmental manipulation. The question may be raised: to what extent will the quantitative measures of single-cell receptive-field properties disclose distributions which justify the classificatory schemes hitherto proposed? In particular, we will examine distributions based on such measures as subfield separation, directionality, contrast independence, and spontaneous activity. In addition to the problem of classification, the criteria which have been used to group cells in striate cortex into the simple, complex, and hypercomplex classes have been under some debate (3, 11). It appears that a portion of the cells considered as simple by some investigators may not have been so designated by Hubel and Wiesel (8) who were the first to describe and name these cells. Almost unavoidably, perhaps, each investigating group has developed its own set of criteria. Since we may too have done so, we decided to call cells in the monkey which we believed to be simple, S-type cells. Those which we thought analogous to complex cells we called CX-type cells. We found that the most reliable classification was achieved on the basis of the spatial location of responses to light increment and light decrement. For this reason we defined our S-type cells, as did Hubel and Wiesel (8), as those orientation-selective cells whose receptive fields had one or more distinct subfields within each of which a flashing stimulus or a stimulus moving in one direction elicited a response to either the increment or decrement of light, but not to both. CX-type cells were defined as those
AND
MONKEY
STRIATE
In two animals one eye was MONKEYS. surgically immobilized by transection of the 3rd, 4th, and 6th cranial nerves. Skull screws and recording wells were implanted prior to recording as previously described (17). During recording sessions the head was restrained, permitting the study of cortical receptive fields via the immobilized eye. The prime purpose of this study was to determine whether or not the receptivefield properties of single cells in visual cortex of the alert animal in our situation were comparable to those of the paralyzed, anesthetized animals. ALERT
Stimulus delivery sys terns Two separate stimulus delivery systems were used, one manual and one computer operated. MANUAL SYSTEM. The manual system consisted of an optic bench, a control box, and assorted power supplies. The optic bench had the following elements: a) tungsten ribbon filament lamp; b) condenser;c) motor-driven aperture permitting adjustment of stimulus length, width, and orientation; d) projection lens; e) X-Y mirror galvanometer. The image was projected to a tangent screen 57.3 inches from the animal. The control box contained a joy stick, which operated the X-Y galvanometers, and several switches which activated the aperture motors. The manual system was used to localize the receptive field on the tangent screen and to determine qualitatively the basic attributes of receptive fields such as orientation, binocularity, directionality, and the location of subfieldsin the S-type cells.
I
1291
bench were driven with stepping motors activated either manuallyor by a PDP- 1l/20 computer. The aperture in this case consistedof two sets of leaves driven by two stepping motors which could produce rectangularstimuli. Stimuli could be made to increase or decreasein size either symmetrically with respect to the center of the optical axis or asymmetrically by keeping one leaf fixed at the center of the optical axis. The next element on the optic bench was the projection lens which wasfollowed by a shutter. After the shutter a right-angle prism permitted the movement of the stimulususingeither a galvanometer or a steppingmotor. The galvanometer was used in conjunction with a waveform generator for smooth stimulusmovement on the tangent screen. The steppingmotor was usedto place stimuli into various fixed positions on the screen. Following this was a dove prism which was rotated with a stepping motor to produce changesin stimulusorientation and direction of motion. All the motor-driven elementsin this system could be manipulated either manually or by computer. The PDP- 11was used and was programmedto allow the manipulation of each of several stimulus dimensions such as stimulus orientation, stimuluslength, and stimuluswidth. One of the major features of the computer program we used, which was developed by A. Polit, wasthat it enabledusto presentstimulus conditions in randomizedorder. The determination of orientation specificity, for example, was accomplished by having a bar or edge move acrossthe receptive field in a numberof different orientations, all of which were presentedautomatically in a randomorder after specificationof the rangeand number of trials. For someof our quantitative measuresthis feature was deemed essentialdue to the inherent responsevariability of single cells in visual cortex. Our data (16) show that the average short-term variability of cortical neuronsto repeated presentationsof an optimal stimulusis 35.4 using the formula: standard deviation/mean x 100. On collecting the data for a given run, the results were displayed on a Tektronix 611 storage oscilloscope; total, mean, and standarddeviation of the numberof spikesobtained to each stimulus presentation could be displayed on command. Thesedata could also be printed out on the Teletype or stored on DEC tape. The stimuli used were 0.8-l .6 log units above cell response threshold. Background illumination was 0.8- 1.2 cd/m?
SYSTEM. Stimulusde- PROCEDURE livery for detailed analysiswas accomplishedby Each unit was first studied by manualoperaa second optic bench. The elements on this tion of the first optic bench. The response COMPUTER-CONTROLLED
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
and subsequently animals were maintained on a 70%-30% mixture of N20 and OZ. Following application of hyoscine, contact lenses of proper curvature were applied to bring the eye in focus on a tangent screen 57.3 inches away. Using a reversible ophthalmoscope the fovea of each eye was projected onto the screen. In most experiments 3-mm artificial pupils were placed in front of the eyes. Animals were typically studied over a period of 2 days. During the night contact lenses were removed for a few hours to reduce cornea1 clouding. Even so, in some animals slight opacity was observed by the end of the 2nd day of recording. When this occurred the cornea1 epithelium was removed by scraping, which resulted in clear optic media (18). Single cells were recorded using glass-coated platinum-iridium microelectrodes (2 1).
CORTEX
SCHILLER,
1292
characteristics quantitatively rangement.
FINLAY,
were subsequently assessed with the computer-driven ar-
Manual assessment
Computer-driven
assessment
The following procedures were used to assess the response characteristics of single cells quantitatively. Each step described below was carried out as a single uninterrupted sequence, initiated by the necessary commands via a Teletype* ORIENTATION. Most commonly we began the detailed study of a cell by initiating a computer command which resulted in having a bar or edge sweep across the receptive field in different orientations in a randomized sequence. The results obtained in this fashion will be described in detail in the second paper. On completion of an orientation sequence the data were stored and displayed on a 611 storage oscilloscope enabling us to see the optimal orientation and the sharpness of orientation tuning. STIMULUS LENGTH. For this series the stimulus was set to the orientation which provided the best response during the orientation sequence described above. The length of the stimulus was varied in a random order, typically over eight steps using stimuli of fixed widths and varying lengths ofO.1, 0.2, 0.4, 0.8, 1.6, 3.2, 6.4”, with 10 or 20 measures in all at each step. Included in this series was a “no stimulus” condition used to obtain a measure of the response rate in the absence of a visual stimulus. MOVING EDGES. To assess the response of units to light (light increment) and dark (light decrement) edges, to determine the spatial location of the response areas, and to assess degree of directionality, single edges or rectangles were swept across the receptive field. Commonly 1” by 1” stimuli were used and most units were tested with both a light square on a dark background and a dark square on a light background.
VOLMAN
These stimuli were moved back and forth across the receptive field in the optimal orientation, usually over a range of 3” at a velocity of 2”/s, for 30 repeated trials. For units with large fields or complicated responses, single edges were used and for strongly “stopped” cells, stimuli with lengths shorter than 1” were introduced. During data collection the responses were displayed as they occurred on the 611 storage oscilloscope in the form of a raster, with each response generating one dot and successive stimulus sweeps appearing on successive lines of the oscilloscope. At the termination of the sequence the data were stored and displayed in the form of a cumulative-response histogram; 256 bins were used, which for 3” back-and-forth movement at 2”/s meant 11.7 ms/bin or 0.023’ spatial resolution per bin. FLASH ES. The responses of cells to stationary flashes were studied using two procedures. One was a single sequence of 30 repeated flashes of an optimal stimulus in the best location with a 1.5 s on and a 1.5 s off cycle. For the second method, several spatially separate sites were flashed. The stimulus was typically a O.09”-wide bar in the optimal orientation of the receptive field. Stimuli were flashed over a range of either 0.8” or 1.6, depending on receptivefield size, using eight different equally spaced positions; 20 trials, with a 1 s on, 1 s off cycle, were obtained for each position in random order. STATIONARY
RESULTS
The results reported in this section are divided into five parts. In the first we will describe the response of various kinds of cells in monkey striate cortex to stationary and moving edges. The second section will deal with the size and shape of receptive fields. In the third section the distributions of cells according to a variety of quantitative measures will be presented. In the fourth section the relationship between unit responses and the site of recording in the cortical layers will be analyzed. The last section briefly describes results obtained in alert animals.
Examples of cell properties assessedby responses to stationary and moving stimuli We defined S-type cells as those oriented cells whose receptive fields had one or more spatially distinct regions within each of which a movingedge or a flashing stimulus elicited a response to either light increment or light decrement. By contrast, CX-type cells were defined as those oriented cells which to either moving edges or stationary flashes of light responded to both light increment and light decrement throughout their
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
When a single cell was clearly isolated and properly triggered by the Schmitt trigger, we began by determining: I) the location and size of the receptive field; 2) the dominant eye; 3) the best orientation; 4) the optimal direction of movement; 5) the responsiveness to light and dark edges or bars; 6) the effect of different stimulus lengths and widths; and in some cases, 7) the response to stimulus velocity within a range of O.S-50%; and 8) the response to colors using red, green, blue, and yellow filters. Some of these measures were again reassessed manually following the computer-operated sequence.
AND
MONKEY
STRIATE
When a single photocell respondstransiently to any sudden change in illumination,
a light or
dark rectangle sweepingacross it will generate four pulsesin a histogram(Fig. IA), two in each direction of movement, one for the light edge and the other for the dark edge. If the photocell circuit is arranged to produce a pulse to light increment alone, only those peaks will be seen which
are labeled
L. If, furthermore,
I
we have
a
FIG. 1. Photocell responses to the light and dark edges of moving rectangles. A: light rectangle on dark background, single photocell responding to both light and dark edges. B: light rectangle on dark background, two spatially displaced photocells, one responding to the light edge and the other to the dark edge. C: dark rectangle on light background, photocell arrangement same as B. Direction of movement across photocells is shown by arrows; turnaround point is indicated by triangle. L, response to light edge; D, response to dark edge. Time is left to right throughout. To facilitate comparisons, thin vertical lines are drawn for each edge centered on responses appearing in the first histogram.
I
1293
the photocell respond to only one direction of movement, only one peak would be seen. Not so obvious is what happens when we place two photocells sideby sidewith somespatial separation between them and have the first respond transiently only to light increment and the second only to light decrement. The responses elicited by moving white and black rectangles traversing the two photocells are shown
in Fig. 1B and C. As can be seenthe pulsesare now asymmetrically displaced with opposite asymmetriesfor the light and dark squares. If responselatency and the total distancetraversed by the moving stimuli are known, the relative location of the responseareascan be calculated on the basisof the temporal separationbetween the pulses(1). CELLS. Our findings suggestthat S-type cells with a variety of responsecharacteristics can be distinguishedin area 17 of the monkey. From the nature of their responsesseveral inferencesmay be maderegardingthe organization of visual cortex. Qualitatively we could discern sevendistinct cell types from our total sampleof 245 S-type cells. An example of each cell type will be shown. Figure 2 showsthe simplestresponsewe obtained to moving stimuli from S-type cells. This cell dischargedonly to a dark edgetraversing the field and did so only in responseto one direction of movement. The stimuluswas moved at right anglesto the axis of orientation so asto elicit the optima1response.Figure 2,1 showsthe response to a white, lo square. In Fig. 2,2, stimuluscontrast is reversed; a black 1” square is swept across the receptive field. In both cases responseoccurs to light decrement, that is, to the dark edgeonly. In the drawing of this figure and subsequentones, the original 256 bins in the computer display were reduced to 128by combining adjacent pairs of bins. In Fig. 2,3 the responseproperty of this cell to moving stimuli is drawn in schematicmanner. It showsthe approximate size of the field and the kind of responseelicited. The arrow indicates that the responsewas unidirectional. This unit may be describedthen asa unidirectional S-type cell, sensitive to a dark edge. This cell respondedvigorously to the turning off of a thin rectangular light bar flashingin the center of the field, as shownin Fig. 3,1. Flashingthis stimulus anywhere but in the center of the field produced no responseat all. A large rectanglecentered on the field, when flashed, also failed to elicit any S-TYPE
response.
However,
when
the center-activat-
ing region was continuously illuminated, the superimposedflashing of a large rectangle did evoke a responseto its onset. These observa-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
receptive fields. For these cells, direction selectivity was not affected by the contrast of the moving stimulus. The results described in this section were obtained by I) sweeping light and dark rectangles or single edges across the receptive field, and 2) flashing stationary targets into various parts of this field. The use of both light figures on a dark background and dark figures on a light background was considered an important reciprocal control for moving stimuli, and the data shown always include examples of both. In most cells a close correspondence was found between receptive-field maps obtained with small stationary, flashing spots and with moving stimuli. As an introduction to the data obtained with moving rectangles, the first figure describes what such responses would look like when the stimuli are swept across photocells attached to the tangent screen where the receptive fields of cells were to be studied.
CORTEX
1294 UNIT
SCHILLER,
FINLAY,
AND VOLMAN 100
78-4-12.30 I
UNIT
-
78-4-12.30
I
W
ON
ON I DEGREES
2 OF
4
3 VISUAL
ANGLE
2. Unidirectional S-type cell located 3” from the center of the fovea which, with moving stimuli, responds only to a dark edge. 1: response to a 1” light square. 2: response to a 1” dark square. Arrows indicate direction of movement across receptive field. Extent of stimulus movement: 3” in each direction, turnaround point indicated by triangle. Velocity: 2”/s. Histogram time is left to right throughout. Vertical lines indicate position on the histogram of the response that a single photocell would give to each edge. The position of these lines has been adjusted for unit response latency. The total number of bins across equals 128, reduced from the original 256 by adding adjacent bins. Number of trials per histogram, 30. L, light-edge response; D, dark-edge response; sp, spikes. 3: schematic drawing showing size and approximate spatiotemporal response properties of the receptive field. Arrow indicates direction of movement across field: the upper part of this figure depicts response to movement in one direction and the lower part response to movement in the opposite direction. FIG.
FIG. 3. Flash response of unit in Fig. 2. Top histogram shows response to a thin bar flashed in the center of the receptive field. Bottom histogram shows response to a large rectangle which was flashed while the center was continuously illuminated. Flashing of this same large rectangle without the small rectangle tonically illuminating the center did not elicit a single response in 30 trials.
activating region of the field. Cells of this sort are commonin area 17of the monkey, and their counterpart, cells responding to a light edge only, are equally numerous. Of the 245 S-type cells studied, 67 were of this type; 34 had light fields and 33 had dark fields. Over 80% of them were unidirectional. We called these S, or single-contrastcells. Figure 4 showsthe responsepropertiesof another cell. It respondedto both a light edgeanda dark edge. The responseswere elicited from two different regions in the field, as is evident from the different locations of the peaks in Fig. 4,l and 4,2. Had they been in the sameplace, the tions suggestthat the cell has a surround which responseswould have centered on the vertical is opposite in sign to the center. This surround lineswhich showwhere a photocelllocatedin the region, however, is unresponsive to moving center of the field would have respondedto each stimuli and to small stationary flashing spots. edge of the stimulus (Fig. IA). This cell is similar to LGN cells in that it hasan Figure 4,3 shows the schematicorganization antagonistic center-surround organization; the of this cell to a moving stimulus, demonstrating surround can be potentiated by illuminating the spatially separate light- and dark-activating recenter region. It is also similar to LGN units in gions. Both of these regions appear to be exthat in its central activating region it responds citatory sincecontrast reversal doesnot alter the only to one sign of contrast. However, in con- responseto each edge. Only the temporal setrast to units in the LGN, this S-type cell is quenceof events changesbecauseof the way the orientation and direction selective, has low stimulusedgesof oppositecontrast move across spontaneousactivity, and dischargesto moving the two subregions(see Fig. 1). The centers of stimuli only while a dark edgemoves acrossthe the two regionsare 0.29” apart. When stationary
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
m
MONKEY
STRIATE
CORTEX
I
1295
UNIT 91-l-22.27
UNIT 72-8-13.55
4 , 0B
OF
VISUAL
z
3
4
5
6
7
9 dep
ANGLE
FIG. 4. Unidirectional &type cell with separate subfields for light- and dark-edge responses. 1: response to a 1” light square. 2: response to a 1” dark square. 3: schematic drawing of receptive field showing size. separation, and directionality of the two subfields. The two subfields responding to opposite contrast respond to the same direction of movement (all parameters as in Fig. 2).
flashes are used a similar spatial separation of the light region (on-response) and dark region (off-response) is revealed. The cell is essentially unidirectional. We called these & or double-field cells; 47 units in our sample had this property. The cell in Fig. 5 shows another commonly occurring property in the cell population of area 17 of the monkey. This cell responds to both light and dark edges, but the direction of the response is affected by contrast. Both the light and the dark edge responses are unidirectional, but they are specific for opposite directions of motion. The locations of these responses are spatially separate, as shown in the schematic spatiotemporal drawing in Fig. 53, where the upper part of the figure shows response to movement in one direction and the lower part, response to movement in the other direction. The spatial separation between the two regions can be derived two ways. It can, first of all, be determined by hand plotting using slowly moving edges or small stationary flashing spots. Second, it can be calculated. This calculation, however, must take into account the response latency of the discharge. This we typically obtained by measuring response latency to flashing
FIG. 5. &-type cell demonstrating interaction between contrast and direction. 1 and 2: responses to I” light and dark squares. 3: schematic drawing of spatiotemporal response. The two subfields responding to opposite contrast also respond to opposite directions of movement.
stimuli of equal intensity. In some cells latency was calculated from responses obtained to different stimulus velocities, as described by Bishop, Coombs, and Henry (1). For this cell, then, the light region is activated by one direction of movement and the dark region by the other. Reversing the sign of contrast between the stimulus-background does not alter this relationship. We called these Z$ or interaction cells because they demonstrate an interaction between direction and contrast. Such units may be thought of as responding best to a single edge moving back and forth across the field which, for this cell, has the light region on the left and the dark region on the right side. When the contrast is reversed, moving the edge within the field elicits no response. When tested in this fashion this is exactly what happens; 36 cells in our sample had this kind of interaction between contrast and direction. Figure 6 shows another common S-type cell. It is one which discharges to a dark edge in both
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
I DEGREES
1296
SCHILLER,
UNIT 84-7-9.15
L-3
FINLAY,
fields. The cells shown in Figs. 4, 5, and 6 have similar spatial arrangements.The directionality of the responseamongthesecells is quite different, however, and is not predictable from the maps obtained with the stationary flashing stimuli. The schematic diagram showsthe spatial arrangement and relative responseamplitudes of the regions of this field. The light area is unidirectional. The dark area may be thought of as a singlebidirectional subregionor as two subregions, one respondingin one direction and the other responding in the opposite direction. Complementary cells which have bidirectional light and unidirectional dark areas are also common. We have a total of 31 suchasymmetric cells in our sample, 14bidirectional for light and 17bidirectional for dark. Thesewere called S4or U-B cells, where U standsfor unidirectionality for one contrast and B for bidirectionality for the other contrast. Figure 7 shows a cell which produces a bidirectional response for both light and dark edges. Both of the light areas and both of the dark areasappear to be in the samelocation for the two opposite directions of movement; 2 1 UNIT 92-I-27.15
FIG. 6. &-type cell with unidirectional light edge and bidirectional dark-edge responses. I and 2: responses to light and dark 1” squares. 3: response to a 0.09’ bar flashed in different locations. Open circles are response to onset of light; closed circles show off-response. 4: schematic drawing of receptive field.
directions but to the light edgein only one direction. This occurred in the samemanner whether 3 one moved a light squareon a dark background (6,l) or a dark squareon a light background(6,2) across the field. Figure 6,3 shows the results of flashing a stationary 0.09” light bar in various parts of the receptive field. The light bar was flashed in six different locations 20 times each in random order. The meannumber of responsesare plotted for the first 250msafter the light is turned on and off. These data clearly show the separatespatial FIG. 7. &-type cell with bidirectional light and dark locations of the light (on) and dark (off) response resnonses.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
3
AND VOLMAN
MONKEY
STRIATE
1297
stimuli they could be mistaken for CX-type cells. Using single edges again, Fig. 9 shows a cell in area 17 with a bidirectional response, which, in both directions, shows three activating regions; a central dark edge region flanked by light regions. The first subfield responds well to leftright movement; in the opposite direction this region is barely observable; 17 cells in our sample had multifield properties of this sort, and were called S, or flanked-field cells. The subfields of some of these cells could not be activated with both directions of movement. Thus some cells had a tripartite field in one direction and only a bipartite field in the opposite direction. The cells we have considered up to now had relatively distinct properties. It would be misleading, however, to claim that all S-type cells were as well defined as these. We have found a minority of S-type cells with intermediate characteristics, suggesting that the various Stype cells, inasmuch as they fall into groups, do not form unique categories. Two examples of such cells appear in Fig. 10. The first (Fig. 1OA)
UNIT 59-3-17.85
FIG. 8. !$type cell with four subfields. 1 and 2: responses to moving single edges. 3: schematic drawing showing that light and dark edge subfields are in different locations for opposite directions of movement.
I
b
FIG. 9. Bidirectional, multiple subfield S-type cell. 1 and 2: responses to moving single edges. 3: schematic drawing of receptive field.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
S-type cells had this characteristic with some variability in the location of the fields as a function of direction of movement, and were called &, or B-B cells. The cells shown so far are relatively common in area 17 of the monkey. Figure 8 shows a less common cell type with rather interesting properties. Because of the complexities of the response and the size of the receptive field, the data shown were obtained by moving only a single edge across the field, with the contrast of the light and dark regions reversed for Fig. 8,l and 8,2. The field was also carefully hand plotted to verify the relationship between the light- and dark-activating regions. The schematic drawing in Fig. 8,3 shows two sets of light and dark regions which are located in different places depending on the direction of movement. Five cells of this type were observed in our sample, and were called Ss or double interaction cells because of the two regions of interaction between direction and contrast. Most cells with this kind of spatial arrangement produce both on- and off-responses to flashing spots throughout their receptive fields. For this reason, if they were mapped only with stationary
CORTEX
SCHILLER,
1298
FINLAY,
AND
VOLMAN
A UNIT 91-l-27.50
FIG.
10.
TWO
L
A
0
L
examples of S-type cells with properties which fell in between some of those previously shown.
is one which responds predominantly to one edge; additional fields are evident, however, placing it in between a single-field S-type cell (Fig. 2) and an S-type cell (Fig. 7). In Fig. 1OB an example is given of a cell whose properties fall between an interaction (S,) cell and a cell responsive to both light and dark edges in both directions (S,). For none of the S-type cells we studied was it possible to predict the directionality of the response reliably on the basis of the receptive-field map obtained with stationary flashing spots. Cells which responded to both CX-TYPE CELLS. light and dark edges throughout their activating region and showed no interaction between direction of movement and stimulus contrast were classified as CX type. There were 342 CX-type cells in our total sample of 1,125 units. Figure 11 shows a unidirectional CX-type cell. It has a relatively large receptive field as shown both with moving edges and with flashing
stationary targets (Fig. 11,3). The data and the schematic drawing demonstrate overlapping light and dark regions. Figure 12 shows a similar cell although with low spontaneous activity and a smaller field. This unit is bidirectional with superimposed light and dark regions in both directions. The two cells in Figs. 11 and 12 conform to the typical complex cells so far described. The majority of CX-type cells are of this sort. We have found, however, that a number of these cells had rather small fields. Such a cell is shown in Fig. 13. By our criteria this is a CX-type cell, yet its field is no larger than that of S-type cells in the same region of the visual field. A small degree of interaction between contrast and direction, typical of &-type (interaction) cells, is also evident in this figure. INHIBITORY
RESPONSES
OF S-TYPE
AND
CX-TYPE
In the cells described so far the schematic drawings were derived from the prinCELLS.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
0
MONKEY
STRIATE
CORTEX
I
1299
UNIT 84-3-13.05
. I
, .2
. .3
. .4
. .s
. .6
.2
.?I
.4
.s
.6
FIG.
11.
. .7
6
.8
Unidirectional
cipal excitatory responses to increase or decreasein energy level produced by either moving edgesor stationary flashing stimuli. In this section we will deal with the inhibitory responsesin S-type and CX-type cellsas observed on the basis of a decrease in responsiveness, which presumably reflects inhibitory action. Bishop, Coombs, and Henry (2) have shown that the inhibitory side bands of simple cells in cat’s visual cortex, in contrast to the excitatory regions, can be mapped when the base-line responseof these cells is increased by repeated stimulation of the excitatory field(s), while at the sametime a moving stimulusis swept acrossthe entire receptive field. This second stimulus inhibits the responseof the cell when it crossesthe inhibitory regions. Similar results have beenobtained by Watkins et al. (20) who, instead of stimulating in the center of the field, increased
.
. .9
9
I 1.0
. 1.1
1.0
I.1
. 1.2
1.3
4 dbg
CX-type cell.
the baseline firing rate of these normally silent cells by injecting potassiumions near the cell. The easiest way to observe such inhibitory responsesis to take advantage of the relatively high degree of naturally occurring spontaneous activity that one can occasionally find in these cells. One S-type cell of this sort is shownin Fig. 14A. This cell had two excitatory subfieldsfor each direction of movement (S, type), similar to the cell shown in Fig. 7. Stimulus-responsehistograms are shown to two different stimuli. In Fig. MA ,I the responsewas elicited with 0.63’ wide bar moving across the field at 2”/s in one direction. Two excitatory responses can be seen, one to the light and the other to the dark edge. Inhibition is evident before and after each response.In Fig. 14A,2 the bar was reduced to 0.27Oin width, resulting in a summationof the light and dark edge responses.The inhibition to
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
,
SCHILLER,
1300
UNIT 85-8-6.45
I
VOLMAN
+--!I--
0
-
240_
% a
160-
Iz3z -0 80-
0’ ,
,
. I
.2
. 3
.
9
.
.4
5
.6
‘0
d 7
de9
4 6
7 dtg
FIG. 13. Bidirectional CX-type cell with small spatially overlapping light and dark subfields showing a small degree of interaction between direction and contrast.
showsonly excitation to light and dark edges.In the opposite direction, however, a smalldegree of inhibition is evident when the stimulusmoves either side of the excitatory areasis even more over the center of the field. This kind of inhibipronounced. tion was seenin many directional CX-type cells. A detailed study of this cell, using various Another group of cells, which stimuli, yielded the schematicdrawing shown in TONIC CELLS. Fig. 14A,3. It appearsthat each region has its we believe forms a rather distinct third classin own inhibitory side bands. While it was clear monkey striate cortex, comprisesthose neurons that the excitatory light and dark areaswere not which are excited by one sign of contrast and in the same place for the two directions of are inhibited by its opposite. Thesecells, which movement, it was not really possible to deter- were called T-type cells, respondto changesin mine whether this was also true for the inhib- illumination in a relatively sustained(tonic) fashitory side bands. In the S-type cells which we ion and the responseis not contingent on having tested in this manner or with Bishop’s method a contour within the receptive field. Typically (2), such inhibitory regions were frequently evi- they are poorly or not at all oriented, and some fire preferentially to colors without beingsharply dent. Similar examination of CX-type cells failed to selective. They appear to be similar to LGN discloseanalogousinhibitory side bands. Figure cells, but someof them are binocular. Two examplesare shown in Fig. 15. The first 14B shows a direction-selective CX-type cell with moderate spontaneousactivity. Movement of these(15A) is excited by light and inhibited by from left to right, as representedin this figure, dark. The opposite is true for the second cell FIG.
12.
Bidirectional
CX-type cell.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
w x a m
AND
UNIT 90-l-21.60
I
320
3
FINLAY,
MONKEY
UNIT
76-3-21.
STRIATE
CORTEX
I
1301
I5
A
(Fig. 15B); 49 cells in our sample were of this type*
Only a small percentage SELECTIVITY. of the cortical units in our sample were color selective. The characteristics of such cells were similar to those described recently by Dow (5) and earlier by Hubel and Wiesel (lo), and hence were not studiedin detail. Most of them could be classifiedas S-, CX-, or T-type cells, although color-coded S-type cells were especially rare. It was our impression that color-selective cells occur in batches; when we found one in a penetration, succeeding cells were similarly color selective. It is, therefore, possiblethat color is representedin a columnar fashion in striate cortex. COLOR
Shape of receptive field and response to stimuli of various lengths We employed a variety of methodsto assess the size of single-cell receptive fields. When
stationary flashingspotswere used, the majority of cells, including S type, yielded roughly circular or elliptical fields, the long axis of which did not necessarilyfall along the axis of orientation. Receptive-field size can also be plotted with moving stimuli. This works well with movement at right anglesto the axis of orientation. If the stimulus moves at less than about 3”/s, the cell will fire as long as the edgeis within the activating region. Hence the size of the responseis a good index of receptive-field size along this dimension,which we will refer to as the width of the field. Plotting the size of the field along the axis of orientation, which will be referred to as its length, is more difficult with moving stimuli. This may be done by moving an optimally oriented edge(at right anglesto the axis of orientation) progressively toward the center of the field from the periphery. Under theseconditions it is not uncommon that the edge has to be moved beyond the center of the field from either side. Thus one may find that having marked the
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
FIG. 14. Excitatory and inhibitory responses of an S-type and a CX-type cell. A: bidirectional S-type cell with two subfields in each direction. A, 1: response for one direction of movement to a 0.63” wide bar. A,2: response to a 0.27” bar showing a summation of excitatory light and dark regions. A,3: schematic drawing of receptive field. Stipled area shows maintained activity. The response profile of subfields to light and dark edges is shown for both directions of movement. B; unidirectional CX-type cell. B, 1 and 2; responses to single edges in both directions of movement, B,3: schematic drawing of receptive field.
1302
A
UNIT
SCHILLER
., FINLAY,
60-2-11.40 I
UNIT
76-3-21.25
I.
FIG. 15. Two T-type cells. A : excitation to light increment. B: excitation to light decrement.
location of the upper border of the field on the tangent screen as the stimulus was advanced beyond the marked point, creating what mi ht be called a “negative field.” What this probaEly meansis that a certain degree of summationis required along the axis of orientationof the field before a responsecan be elicited (3). Another approach for measuringthe length of the receptive field is to increasethe length of a stimuluscentered in the field in small stepsuntil the responseasymptotes or begins to decline. The size of the stimulusat this point should be equivalent to the length of the activating region. This methodalsoallowsoneto determineto what extent a unit might be inhibited as the length of the stimulus is progressively increased. Such a measurehasoften been usedto place cells in the hypercomplex category (9). In the cat’s and
VOLMAN
monkey’s visual cortex these cells have been shownto have inhibitory regionsalongthe axis of orientation so that once a stimulus had been increasedbeyond a certain lengththe responseis attenuated. Units with such “end stopping” are classedas hypercomplex cells. Simpleand complex cellsin the cat arenot inhibitedin this manner (9), but their summationproperties to stimuli of increasinglengthsappearto be different. Simple cellshave beenshownto summatewhile complex cells are less apt to do so. In order to obtain a measureof receptivefield length, summation, and end stopping, we moved optimally oriented slits of various lengths acrossthe receptive fields of single cells. Most commonly the stimuluslengthsusedwere 0,O. 1, 0.2,0.4,0.8, 1.6,3.2, and 6.4Oof visual angle,the first beinga control no stimuluscondition. These stimuli were swept acrossthe receptive field in a randomized order for 10or 20 trials per stimulus condition utilizing our computer-driven optic display. Stimulus velocity was typically around 2”ls. The results for a strongly stopped unit are shownin Fig. 16in the form of a setof histograms and a curve for one direction of stimulusmovement. This unit had low spontaneousactivity. Responsesrapidly summated with increasing stimuluslength reachinga maximumwith the 0.8 stimulus. With further increases in stimulus length, the responsedeclinedsothat the 6.4”-long stimulus elicited only 15% of the maximum response.This ratio can be usedasan index of end stoppagefor each cell: stopping index = 100 - 6.4” stimulus response x loo best response
Figure 17 shows a representative sampleof nine S-type and nine CX-type cellsstudiedin this fashion from a sampleof 572 units. All the cells shown in the left-hand column of this figure had spatially separate light and dark regions and were, therefore, classifiedas S-type cells. Cells classifiedas CX type had overlapping light and dark regions. We wish to make the following points about this figure: I) The slopesof summation among both S-type and CX-type cells are highly variable and show no consistent differencesfor the two types. Thus, thesecells in the monkey do not seem to differ dramatically in terms of summation. 2) There is considerable variation in the extent to which cells are stopped (index of stoppage at right extreme of each graph). S-type and CX-type cells do not differ in this respect. The relatively continuous variation in the degree of inhibition among these cells suggeststhat the criterion of end stopping may
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
B
AND
MONKEY
CORTEX
I
1303
60-8-12.35
IOSP
-0,?
0 Ids
.I
.
.
.
.2
.4
.8
1
.
I 6
3.2
I
6.4
.I Q STIMULUS
LENGTH
IN
DEGREES
FIG. 16. Response of a strongly stopped CX-type unit with a receptive-field width of 1” to stimuli of increasing length. Histograms are for one direction of movement, reduced from 128 to 32 bins.
not adequately distinguish a separate class of undertaken in an area representinga relatively hypercomplex cellsin the monkey striate cortex. narrow range of the visual field, spanning2”-5” We will examine this question in more detail in from the fovea. Figure 18showsthe distribution the next section. 3) The maximum responseoc- of overall receptive-field widths in this region. curs anywhere from 0.2’ to 6.4”. This would The width of the field was obtained with stimuli suggestthat the length of the receptive field is moving at right anglesto the axis of orientation. often indeed greater than its width, a finding The duration of the moving edge responsewas which does not correspond with that obtained measuredat the half-height point between the using other indexes of receptive-field length. base line and the peak of the response(area containing approximately 75% of the response). Distributions based on For cells having spatially separateresponserequantitative measures gions, field width was measuredto include the The data describedsofar have beenprimarily entire area these regions covered. To assessthe validity of this method we comin the form of individual examples of cell responses. In order to assessthe validity of pared the receptive-field width of 50 units using classificatory schemesand to obtain a general two different measures;the one just described picture of single-cellfunction in visual cortex of andone derived on the basisof flashinga 0.09’ bar the monkey, we examined the distribution of in various parts of the receptive field usingthe computer-driven display. These two measures these cell properties in a large sample. Table 1gives an overall account of the number yielded similar results, showinga correlation of of cells studied and the gross categories into +0.69. This relationship would probably have which they were placed. The criteria used for beeneven better had we beenable to activate the classificationof S-type cellsappear in the legend field reliably with thinner bars, thereby increasof this table. These criteria permitted us to ing the sensitivity of the flash measure. The classify all but 21 S-type cells. This suggeststhat receptive-field sizesderived usingmoving stimuli the cellswe distinguisheddo not form continuous also correspondedclosely with our hand plots. The data in Fig. 18, obtained from 589 cells, distributions in their properties asalready noted. We will next turn to a number of quantitative show a skewed distribution of sizes which vary measuresto determine to what extent the basic over almost a 20-fold range. Only a relatively criteria which had been used in classifying cells smallfraction of this rangeis due to variation in might be applicable for single cells in monkey retinal eccentricity. Even in a vertical penetration, where receptive fields scatter only over a striate cortex. smallarea, field size can vary over a large range, RECEPTIVE-FIELD WIDTH. Our recordingswere with the receptive fields of cells in layer 4 being
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
UNIT
STRIATE
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
MONKEY TABLE
1.
Account
STRIATE
CORTEX
of cells studied
1305
I 127
TOTAL
Total
N 8589
S-type cells CX-type cells T-type cells Unoriented cells Unresponsive cells Unclassified
245 342 49 62 130 297
Total
1,125 S cell
breakdown
1. Response to one edge a. Unidirectional, L only b. Unidirectional, D only c. Bidirectional, L only d. Bidirectional, D only 2. Response to two edges a. Unidirectional b. Bidirectional c. L in one direction, D in d. L in both directions, D e. D in both directions, L 3. Multiple responses a. Three fields b. Two opposite direction 4. Uncategorized
28 28 6 5
.4
.5
.6
.7
xl
.9
IO
II
1.3
14
I.5
1.6
I7
I.8
I9
2.0
I.6
I.7
I.6
I.9
2 0
1.6
I7
I.6
I.9
2.0
I.7
I.6
I.9
20
ce IIS
type N=l60
47 21 36 14 17
other in one in one
12
17 5 21
fields
Total
245
Criteria for S-type cell classification (calculations for these criteria are described in RESULTS, Distribution bused on quantitative measures. 1: contrast measure is less than 15 (response to dark edge only) or more than 85 (response to light edge only) (S, type). 2a: two separate subfields where 15 < C < 85, interaction index is greater than 10, overall directionality is less than 50, and the difference between directionality for the light edge ( Di) and that for the dark edge ( DD) is less than 50 (S, type). 2b: Two separate subfields, 15 < C < 85, I > 10, overall directionality more than 50, and 14. - D,,I < 50. (S, type). 2c: I s 10, 15 < C < 85, DL and DD are both G 30 in opposite directions (S, type). 2d, e: 15 < C < 85, I > 10 and 1 D1, - DI,I s 50 or DL(DD) > 30 if Q,( D,) < 10 (S4 type). 3a: Three subfields alternating light and dark in at least one direction (S, type). 3b: Two separate subfields where each meets the criteria for 2c (& type). Unclassified: not enough data to classify but enough separation of subfields to be called S-type cells.
cx
type
cells
N=262
.8
IO
-
.9
IO
I I
I2
I3
I4
Unotiented
9
IS
cells
N=39
.I
.2
.3
.4
.S
.6
.7
.B
.9
I.0
RECEPTIVE
FIG.
18.
Distribution
I.1 FIELD
of
I2
I3
I4
IS
1.6
SIZE
overall
receptive-field
width.
S-type and CX-type cells can be differentiated. This can be expressed in terms of a potential misclassification between these classes. Choosing the best criterion to differentiate S-type from CX-type cells on the basis of subfield size results in a misclassification of 18% of the population. Using the same criterion for overall receptivefield size, 34% of the cells are misclassified. SPATIAL
SEPARATION
BETWEEN
SUBFIELDS.
CX-type cells and many S-type cells respond to both light and dark edges. By defini-
tion, the prime distinction between these classes of cells is that the subfields of S-type cells are spatially separate, while those of CX-type cells are spatially superimposed. To check on the validity and the classificatory power of this distinction we compared the extent of spatial overlap between the light and dark subfields in S-type cells with bipartite fields and in CX-type cells. This was done by using the measure of receptive-field size and the center to center spatial separation between the subfields. The follow-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
1. 2. 3. 4. 5. 6.
distribution
1306
SCHILLER,
FINLAY,
67
s type N=
type
cells
N=258
FIG. 19. Distribution of subfield widths for S-type and CX-type cells. The light-edge receptive field was used unless the dark-edge response was better.
ing formula was used to determine the extent of separation or overlap between the subfields: overlap = 1/2L + 1/2D + center-to-center l/2L + 1/2D - center-to-center
field separation field separation
where L standsfor the receptive-field size of the light-edge responseregion and D standsfor the receptive-field size of the dark-edge region. A
value of 100obtained this way is equivalent to a 100%overlap betweenthe light anddark fields. A value of 0 indicates that the two subregionsare immediately adjacent without overlapping. A negative value indicates a space between the subfieldsas determined by this measure. Figure 20A shows the comparison between S-type cells with bipartite fields and CX-type cells. Thesedata showthat our criterion successfully discriminatesthese two types of cells, with only a few falling into a common area. A few S-type cells had subfieldswith an unresponsive region betweenthem. This is somewhatexaggerated in Fig. 20 becausereceptive-field size was measuredso as to include only 75% of the responsearea. For example, the value for the cell in Fig. 4 was -18 although it is clear that there is indeed a small degree of overlap between the subfields. By changing the receptive-field size criterion to the less reliable but more encompassinglevel of 90% of the responsearea, one would effectively shift the 0 point for S-type cells to the -20 region on the figure. This still leaves a number of cells which show an unresponsive area between the subfields. These data suggestthat the prime criterion usedin differentiating S-type and CX-type cellsis an effective one. Our analysisshowsthat choosing the best criterion to differentiate S-type from CX-type cells on the basisof subfieldseparation results in misclassification of only 4% of the B
A
60 -
s type cells N=ll2
?
SUBFIELD
SEPARATION
z 3 IA 0
~
20-
: r z 40.
CX type cells N=l84 60 -
FIG. 20. A: spatial overlap between light- and dark-edge response areas for bipartite S-type and CX-type cells. B: center-to-center separation of subfields of bipartite S-type cells.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
CX
cells 144
AND VOLMAN
MONKEY
STRIATE ICO
1307
I
143
TOTAL N=762 00 -
20
m 60t z 3
30
40
!iO
60
70
8-O
90
100
50
S type cells
IO
20
30
40
50
60
70
80
90
100
CX type
cells
N= 313
DIRECTIONALITY. To determine the distribution of direction specificity, for each cell a count was made of the total number of responseselicited by moving edgesin each direction to both signsof contrast basedon 60 trials. An index of direction specificity wascalculatedby the following formula:
index of directionality
= ;z;
Ex
loo
where the smallernumber was always placed in the numerator. This measuredoes not reflect anything about absolute direction in the visual field but, instead, specifiesthe directionality of a cell to opposite directions of movement at right anglesto the axis of orientation. A small value reflects a tendency for the cell to respondonly in one direction of movement; such cells are thus consideredpredominantly unidirectional. A high value, such as 90-100, indicates that the cell respondedequally well to both directions and, hence, could be called a bidirectional cell. Figure 21 shows the distribution of direction specificity for 762 cells, of which 220 were classified as S type and 3 13 as CX type. The CX-type cells show a roughly bimodal distribution, suggestingthat unidirectional and bidirectional cells may be considered as two separate subpopulationsof CX-type cells. The distribution for S-type cells is skewed, strongly favoring unidirectionality. Sincethis sampleincludescells which show an interaction between responseto direction and to contrast (S, type), we attempted
IO
U
20
30 INDEX
40
50 OF
60
70
80
90
100
B
DIRECTONALITY
FIG. 21. Distribution of directionality. The index of directionality is 100 times the total response in the worse direction divided by the total response in the better direction. An index of 100 indicates a bidirectional cell; an index of 0 indicates a unidirectional cell.
to analyze the extent of directionality by reducing this confounding variable. We did this in two ways. First, we analyzed all single-contrast,S,type cells separately. Figure 22A shows this analysisfrom which it is clear that thesecellsare predominantly unidirectional. Second, we determined for all S-type cells the extent to which either a light or dark edgealoneelicits a unidirectional response.The distribution derived in this manneris given in Fig. 22B. It showsthat when the interaction between contrast and direction is reduced asa variable, the unidirectional properties of S-type cells are more apparent. This kind of analysisdoesnot affect the distibution of complex cells as given in the previous figure because thesecells (seealso Fig. 24) showno interaction between contrast and direction. .AST. To assesshe overall selectivity of striate cellsfor contrast we counted the number CONTR
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
population. The overall differentiation of these two classesof cells may actually be even better, since this sample does not include the easily categorized single-contrast, S-type cells. In Fig. 20B we plotted the center-to-center subfieldseparationof S-type cellsin degrees.The majority of cells had 0.2-0.3Oof center-to-center separation. The average spatial separation is about as great as the average receptive-field size of each subregion(seeFig. 19).CX-type cellsare not shownin Fig. 20because,with the exception of three cells, they would all falI into the O-O.1 column. We were also interested in determining for S-type cells the relationship between the size of the subfieldsand the degree of center-to-center separationbetweenthem (measuredin degreesof visual angle). A correlation performed between thesetwo measuresusing 135S-type cellsdid not show a significant relationship(Y = 0.15). Thus it appearsthat in our sample,which was restricted to a region 2”-5” from the fovea, the size of the subfieldsin bipartite field S-type cells and their spatial separationsvary independently of each other.
CORTEX
1308
SCHILLER,
FINLAY,
AND VOLMAN
TOTAL 160
-
N= 654
20
30
40
50
60
70
80
90
100 10
20
30
40
50
60
70
80
90
100
S type
cells
N=213
IO
20
30
40
50
60
70
80
90
100
W z 3 L
124 I20
-
100
-
80
-
60
-
CX type
IO
20 INDEX
30
40 OF
50
60
70
80
90
100
DIRECTIONALITY
Index of directionality for S-type cell subfields. A: units which respond to only one sign of contrast. B: directionality index distribution of single contrast subfields considered separately. The proportion of unidirectional fields increases when light and dark response subfields are considered separately. FIG.
cells
N=308
22.
of responses elicited by the light and dark edges separately, independent of direction of motion. These data were obtained, as described earlier (Figs. l-13), by moving rectangular stimuli or single edges across the receptive field at right angles to the axis of orientation. An index of the contrast specificity of a cell was derived by the following formula: contrast specificity = XL+CL CD ’ loo where L and D stand for the responses elicited by the light and dark edges. An index value of 10 or less means that the cell responded almost exclusively to a light edge, while a value of 90 or more indicates a strong preference for dark edges. A score of 50 indicates an equal response strength to light and dark edges. In Fig. 23 the data based on 654 cells show
.O
20
30
40
50
60
70
80
90
IO0
Unoriented
cells
N =50 ‘0
20
30
40
50
60
OF
CONTRAST
I
70
80
90
100 D
INDEX
FIG. 23. Distribution of contrast dependency for total, S-type, CX-type, and unoriented units. The index of contrast = (response to light edge/response to both edges) x 100. An index of 100 indicates response to light only; a unit responding equally to light and dark has an index of 50; a dark response only gives an index of 0.
that the majority of cells in our sample responded fairly equally to light and dark edges. CX-type cells as well as unoriented cells show this tendency most strongly. By contrast, S-type cells are evenly distributed on the scale, with
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
lo 131
MONKEY
STRIATE
CORTEX
1309
I
many cells respondingonly to one sign of contrast. Not included in this sampleare the tonic cells, all of which respondin an excitatory fashion to only one sign of contrast. INTERACTION TION. We
BETWEEN
CONTRAST
AND
DIREC-
100
m Ir 3 L 0
a w
-
s type N
cells
8211
50 -
IO
20
30
40
50
interaction = JL+ + D+1 Or [L+ + D+l x 100 total L + D The smaller combination was used in the numerator. This measure ranges from 0 to 50, where a value near 0 reflects very stronginteraction of the sort shown for unit 91-I-22.27 in Fig. 5, while a value near 50 would indicate no interaction. Figure 24 showsthe distribution of interaction valuesfor 689cells, of which 211were classified as S-type, 301as CX-type, and 50 as unoriented cells. These data show that CX-type cells, as stipulated, do not have a significant degree of interaction. S-type cells show some bimodality in their distribution, which is indicative of two subgroups, one showing a low and the other a high degree of interaction. lNHIBITION LENGTH.
WITH
INCREASING
STIMULUS
The measurefor inhibition along the axis of orientation was obtained by sweeping bars of different lengths across the receptive field, as already described. Stopping index (stimulus length-responseratio) was calculated by dividing the response obtained with the longest stimulus(6.4”) by the optilmalresponse. This fraction was multiplied by 100 and this value was subtracted from 100. If the cell’s maximum response was elicited by the 6.4” stimulus, a value of 0 was assignedto that unit. Thus, a low value on this scale indicates an unstopped unit, that is, one which responds without being inhibited by increasingthe length of the stimulus. A value near 100indicatesa cell which is strongly inhibited by stimulation of regionsalongthe axis of orientation away from the receptive-field center. The distribution for end stoppageshown in Fig. 25 was determinedfrom the analysisof 572 cells, of which 129were S-type, 226-CX-type, 78 INDEX
OF
INTERACTION
FIG. 24. Distribution of interacti on between contrast dependence and directionality. The index of in-
teraction (1) = (L-+ + D+- or L+- + D--+/total L D) x 100. An I of 50 shows no interaction; a unit with an I of 0 responds to light and dark edges in opposite directions.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
have shown that somes-type cells exhibit a strong interaction betweencontrast and direction. To measurethe distribution of this interaction, the samedata, obtained by moving light and dark edges over the receptive field, were reanalyzed to obtain an index of interaction. This was computed for each unit usingthe following formula:
1310
SCHILLER,
60- TOTAL N=572
60-
80
we S type
70
60
50 40
30 20
IO
cells
N = 129
10090 v)
80 70
60
50 40
30 20
IO
60-
+ 5 3
50- CX type cells N=226 40-
100 90 80
70
60
50 40
30 20
IO 21
20-
Unoriented
cells
N=70 IO-
100 SO
80
70
60
50
40
30
20
50 40
were unoriented, and 19were T-type cells. The distribution appears to be primarily unimodal with no clear difference between the classes.It is our impressionon the basisof thesedata that this measure cannot be used to separatecells reliably into a hypercomplex category in the monkey. In order to obtain an idea about the extent to which an analogueto end stoppingis observable in the LGN, we also studied 16cells in the same 2”-5’ area of the visual field using similar methods. The index of stoppagefor thesecells was15, 21,22,23,27,33,38,39,43,45,49,63,65,74,75, and 100.It may be saidthat mostLGN cellsusing this method show pronounced surround inhibition and, hence, demonstrate strong end stopping. In what senseend stopping representsa cortical transformationthen isa problem. We will consider this question in the DISCUSSION. SPONTANEOUS ACTIVITY. The last distributional measuredealswith spontaneousactivity, which in previous work has been shown to be lower for simple cells than for complex onesin the cat (1). The meannumberof dischargesoccurring with no stimulus in the receptive field was measured during the testing of various stimuluslengths. A l-s time period was sampled randomly during these trials. The distribution for spontaneousactivity is given in Fig. 26. It showsthat S-type cells have lower spontaneousactivity than CX-type cells (meanfor S = 1.2, for CX = 4.9). Yet there is a considerableoverlap in population. Particularly evident is the fact that there are many CX-type cells which have low spontaneousfiring rates. Choosingthe bestcriterion to differentiate S-type from CX-type cells on the basisof spontaneous activity results in misclassificationof 27% of the population.
Distribution of recep tive--eld properties of cells in cortical layers Using lesions to mark electrode tracks, we identified the location of 201 cells in striate cortex. Figure 27depictsthe reconstruction of part of a tangential penetration between two marker lesionsin which 18 isolated single cells were recorded. Two cells were not held long enoughto
IO
lo- T type cells
10090 80 70 60 STOPPED STIMULUS LENGH
AND VOLMAN
30 20 IO UNSTOPPED RESPONSE RATIO
FIG. 25. Degree of inhibition as a function of increasing stimulus length. The measure of end stoppage (stimulus-length response ratio) was obtained by sweeping the receptive field with bars or edges of various lengths and applying the following formula: 100 - (response to bestlength/response to longest bar) x 100. Largevaluesindicatestronglystoppedcells.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
10090
FINLAY,
MONKEY
STRIATE
CORTEX
I
1311
edge responseis represented in the schematic drawings. The short, heavy line in the center of each schematic drawing representsthe axis of orientation. The stimuli used to generate the data from which the drawingswere derived were 300 moved acrossthe receptive field at right angles to the axis of orientation, as described earlier. The markingsare in 0.1’ steps. The main points to be consideredabout this figure are the following: I) Both S-type and CX-type cells can be 200 found in layer 6. This is also true for T-type cells. 2) Receptive-field size varies over a broad range amongthe cells in the two layers sampled in this pass.3) Nearby cells have similar orientations (12) but their directions can be 180” reversed. This is especially evident for the cells numbered 15 and 16. These two cells were recorded from simultaneously.4) The double-field &-type cell (7) hasa small center-to-center separation, which with stationary, flashing stimuli + 2 8 16 32 might not have beendiscernible. This cell is only marginally different from a small CX-type cell. 116 This cell falls into the overlapping region of Fig. 20. 5) In this penetration there are four singleS type cells contrast, S-type cells in layer 6 (numbers6, 9, N= 137 15, and 16). Figure 28 showsthe overall distribution of cell types in the cortical layers (11). S-type cells were found in all layers, and in layer 4c most of the isolated cells recorded were S type. This layer containspractically no CX-type cells, which are numerousin all other layers. The samplein Fig. 28is basedon only 201cells and does not include data from a number of penetrations, not reconstructedhistologically, in I6 32 + which we estimated the location of the cells in cortex by noting when the electrode touched brain and when it entered white matter. The CX type cells relative thickness of layers in striate cortex is quite constant, with layers 1,2, and 3 constituting 81 N=245 the top 35% of tissue, layers’ 4a, b, and c, the next 30%, and layers 5 and 6, the remaining25% of gray matter. We could, therefore, assigncells with a reasonabledegree of accuracy to these three subdivisions. The locations of 295 cells were determinedin this way giving a total of 496 cells whose approximate positions in cortex were known. The number of S-type cells in layers l-3, 4, and 5-6 were 29, 36, and 35, re16 32 + 2 8 spectively; there were 62, 48, and 50 CX-type SPIKES PER SECOND cells in these subdivisions of cortex. Of the Distribution of spontaneous activity for single-contrast, S-type cells there were 8 in FIG. 26. total, S-type, and CX-type cells. layers 1-3, 8 in layer 4, and 7 in layers 5-6. T-type cells had the following distribution: be identified. Two were classifiedas tonic types layers l-3, 3; layer 4, 7; layers 5-6, 7. In Fig. 29 distributions among the cortical (T), five as S type (6, 7, 9, 1.5,16) and eight as CX type. The receptive-field organization of layers are shown for receptive-field size, inhibithese cells in terms of optimal orientation, direc- tion along the axis orientation, and spontaneous tionality , receptive-field size, and light-dark activity. 400
-
381
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
u
1312
SCHILLER,
FINLAY,
AND VOLMAN
2 3 40 4b
14
13 I2
II
10987654
3
2
I
11
FIG. 27. Tangential penetration reconstruction (layers S and 6). Filled circles show position of marker lesions. Receptive fields were mapped with moving edges or squares and are shown schematically marked in O.l” steps; the short, heavy lines are the axes of orientation. Cells number 6, 7, 9, 15, and I6 are S type; 5 and I I, T type; all others are CX type. The two units between 10 and II were not held long enough to classify.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
16 I5
MONKEY Number
20
ii )r
CORTEX
1313
I Percent
cells
60
-
40
-
60
-
40
-
20
-
20
-
of
cells
-
-
IO -
0 W
tz W z
IO
-
2-3
40
4b
4c
6 CORTICAL
2-3
40
4b
4c
LAYERS
FIG. 28. Distribution of cell types in cortical layers. Locations of 201 cells were determined from histological reconstructions. Graphs at left show numbers of each type cell found in the various layers. Graphs at right show the percentage of cells in each layer classified into the various cell types.
These results indicate that: I) The receptivefield size of S-type cells is the same in all the layers, except for layer 4c (not shown separately in Fig. 29for which 4a, b, and c were combined), where all the cells had fields smaller than 0.2”. Receptive fields of CX-type cells are smallest in layers 1-3, and get progressively larger in deeper layers. 2) The degree to which cells are inhibited along the axis of orientation varies considerably with depth. Cells in layers 1-3 tend to be more strongly stopped than cells in layers 5 and 6, where most cells are unstopped, while layer 4 falls between these extremes. The differences between S-type and CX-type cells are small. 3) S-type cells have low spontaneous activity in all
layers of cortex. CX-type cells also show low spontaneous activity in layers 1-3, but tend to have higher rates of maintained firing in layers 5-6. This measure seems reasonably effective in discriminating between S-type and CX-type cells in layers 5 and 6, but is ineffective in the other layers. In contrast to the overall misclassification of 27%, in layers 5-6 only 11% of the cells are misclassified using spontaneous activity as a criterion.
Alert monkeys We were concerned in the course of our recordings that the state of the animal, including anesthesia, might produce sufficiently pro-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
20
of
STRIATE
1314
SCHILLER, FIELD
FINLAY,
AND VOLMAN STOP
SIZE
N= 117
SPONTANEOUS
N=62
75-
75-
50-
50-
25-
N= 103
N=225
25-
cn
.2
.4
.6
.8
1.0
t
FlEL,D SIZE
.2
.4
IN DEGREES
.6
.8
1.0
+
100
LAYERS
80
60
40
INDEX
20
loo
80
60
40
20
OF STOPPAGE
I
LAYERS
4
8
I6
SPIKES
t
I
PER
4
8
I6
+
SECOND
FIG. 29. Field size, inhibition along axis of orientation, and spontaneous activity as a function of depth for S-type and CX-type cells. Depths were obtained from reconstructions and the estimation method described in the text. CX-type cells become larger, less stopped, and more spontaneously active with increasing depth; S-type cells show a change only in their stopping index.
nounced changesin striate cortex to render the properties of singlecells in our situation different from those of the alert, normally behaving monkey. Our paralyzed animals were anesthetizedinitially by Pentothaland subsequentlyby N20. We did not seeany striking differencesin the properties of cells using these two agents. The spontaneousactivity and responsivenessof cells was somewhatlesswith Pentothal, but this was most evident only for the first 5-10 min following readministration. Irrespective of the anesthetic agent used, the responsivenessof cells in the superficial layers (but not in layers 4, 5, and 6) tended to decrease with time, especially during the latter part of the second day of recording. Becauseof theseobservationswe undertook to record from two alert monkeys which had one eye surgically immobilized as described in the METHODS section. Only qualitative data were obtained in these animals. Our recordingsfrom 109 cells suggestthat the properties of units in these animalswere quite similar to those of the paralyzed, anesthetized animals. In the superficial layers many cells respondederratically and tendedto habituate, while the cells in layers 4, 5, and 6 dischargedconsistently to visual stimuli. The majority of cells were oriented although this sample seemedto contain a somewhat greater percentage of unoriented cells. We also found fewer S-type cells, which may have been due to the fact that the recording stability and time spent
in a given region was lessthan in the paralyzed animals. In general, our qualitative observations suggestthat the properties of singlecellsin anesthetized, paralyzed animalsare representativeof the alert monkey’s visual cortex. DISCUSSION
This section discussesthe various types of cells described in the RESULTS and the significance of the distributional measures.Possible models for cortical cells will be considered briefly.
Characteristics striate cortex
of single cells in
The resultssuggestthat a variety of S-type cells may be distinguishedin the striate cortex of the monkey; their properties are summarized schematically in Fig. 30. The singlecontrast, S-type cell in this figure shows the simplest type of oriented cell, which exhibits a singleexcitatory region to moving edges,within which it respondsonly to one sign of contrast change. In some of these cells it is possibleto demonstratean antagonistic surroundby placing a stationary stimulusin the center while flashinga large, superimposedstimulus. These cells are striking in that practically all of themare unidirectional. The secondnotable group is the double-field, !$ cell, in which one subfieldis excited by a light edge and the other by a dark edge. Both fields S-TYPECELLS.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
l-3
l-3
MONKEY
2
3
OF
DEGREES
1315
I
4
VISUAL
OF
CORTEX
ANGLE
VISUAL
AhSLE :6
.I
2
.3
4
5
6
7
6
.9
dsQ
S .I
.2
.3
.4
:6
:5
7
6
9
1.0
de9
6 -
O'JQ
I
FIG.
2
3
4
5
6
7
8
9
IO
I I
I
2 , deQ
30. Schematic drawings for seven S-type cells and one CX-type cell.
respond to movement in the same direction, and two-thirds of such cells in our sample were completely unidirectional. The third group are cells which show an interaction between contrast and direction of movement (S,). They have two subfields, and
each responds only ment, but in opposite The fourth group S-type cells are those elicits responses for ment, while the other
to one direction of movedirections to each other. of commonly observed in which one contrast edge both directions of movecontrast edge elicits a re-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
I DEGREES
STRIATE
1316
SCHILLER,
FINLAY,
VOLMAN
mean receptive field of the single-contrast, S, cells in our sample is 0.200 (SD = 0.09); that of the double-field, S, cells is 0.40 (SD = 0.10). The proposed combinations do not go contrary to the known cytoarchitecture of cortex: while orientation is strictly maintained in a columnar fashion, no such organization is evident for direction; within a column reversals of preferred direction are common. Furthermore, one can, on occasion record from two single-contrast, S, -type cells simultaneously, as already noted, which have spatially separate fields and are sensitive to opposite directions of movement. The problem with any sort of hierarchic concept is that recordings from S-type cells, including S-type, are obtained in all cortical layers. One must, however, consider the possibility that the order of cell types in these layers is greater than revealed by extracellular recording methods. If single-unit records were obtained not only from cell bodies, but to some extent also from axons and dendrites, such order would be significantly obscured. In this respect it is interesting that progressive changes in receptive-field properties as a function of recording depth in cortex are found only in CX-type cells. If unit signals are indeed recordable at sites other than near the cell body, the probability of this would be higher for S-type cells provided they have the profuse axonal terminations within striate cortex postulated by the Hubel-Wiesel (8) hierarchic model. The proposed hierarchy is only one possible way of conceptualizing the various S-type cells since models of cells can also be constructed assuming no such hierarchy. A more detailed discussion of models of visual cortex will be deferred until the fifth paper in this series. Irrespective of the manner in which these cells are constructed, the question remains as to how a single cell can have spatially separate excitatory subfields of the sort described. Since input from the LGN seems highly ordered topographically (12), one possibility is that inputs converge from spatially separate cortical areas, from one or two hypercolumns away where cells of the same orientation but of a slightly different spatial location reside. The other possibility is that the different layers of the LGN, consisting of repeated representations of the visual field project into visual cortex with some misalignment. These projections form multiple layers of terminal fields (6) and thus could be the source of the spatially separate subfields observed in S-type cells. If this is the case one might expect that in some cases the subfields have their origin in different layers of the LGN. This notion does not explain one interesting problem, namely that there are potential combi-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
sponse only for one direction of movement. These cells were called U-B or S4 cells. More complicated arrangements are observed in S-type cells which have several subfields (S,, S,, and S,). Some of these cells-may have three distinct areas with, typically, a center region responding to one sign of contrast change flanked by two regions responsive to the opposite contrast (S,). In some cases these cells are completely bidirectional, while others may demonstrate unidirectionality for some regions and bidirectionality for others. The properties of the various S-type cells we discussed are similar to the simple cells described by Bishop, Coombs, and Henry (1) in area 17 of the cat. In Fig. 10 of their paper these investigators show a variety of spatial arrangements of simple-cell subfields. We have seen cells similar to almost all of these. It is rather interesting that this similarity in cat and monkey cells extends, in part, to the percentage of each cell type; in the sample of Bishop et al., 9 of 43 cells responded only to a single edge, and all of these cells were unidirectional. Of the cells responding to 2 edges, 22 were unidirectional and 7 bidirectional, again a ratio rather similar to the one we obtained in the monkey. However, we have seen more cells showing interaction between direction and contrast (S,) of which Bishop, Coombs, and Henry report two. The U-B cells (S,) were also more common in our sample. How might the more elaborate receptive-field arrangements of S-type cells be derived from their inputs? While a variety of possibilities exist, the array of S-type cells described might well be understood by assuming a hierarchic relationship among them. Lowest in the hierarchy are the single-contrast, S, cells, which may be thought of as being first-order S-type cells. Assuming such a hierarchy, the initial transformation of LGN input in visual cortex involves not only orientation, but also direction selectivity: it appears that at this early stage a conversion occurs which renders almost all cells unidirectional. S-type cells with more complicated properties may be constructed by convergent excitatory input from the first-order S, cells. Thus a double-field, & cell may result from two singlecontrast, S, cells of opposite contrast response, but with the same orientation and direction. An S3 cell may be constructed in the same manner, but the two single-contrast, S, cells would have direction preferences 180’ in opposition. Cells with multiple subfields may be made from a variety of combinations. In favor of such a hierarchic relationship among S-type cells is the fact that the receptive-field width of these cells increases with increasing complexity. This is especially clear when one compares the S, - and $,-type cells. The
AND
MONKEY
STRIATE
CELLS. We defined CX-type cells as thoseoriented cells which have a unified activating region within which a responsecan be elicited to both light increment and light decrement. The receptive fields of these cells tend to be larger than those of S-type cells, although there is some overlap in the distribution obtained. The measure of direction selectivity suggeststhat these cells constitute two subgroups: unidirectional and bidirectional cells. CX-TYPE
CELLS. The T-type cells, which appear to form a distinct group, are characterized by the fact that they respondto visualstimuli in a relatively sustainedfashion and are excited by only one sign of contrast. They are generally poorly oriented, or not oriented at all, and a higher percentageof them are color specific than S-type and CX-type cells. Thesecells may be analogous to the class I cells reported by Dow (5). Since their properties are somewhat similar to LGN cells, they do not appearto transform the geniculate input greatly. Therefore, it is likely that these cells are driven primarily by geniculocortical afferents. T-TYPE
Quantitative
measures
The basic criteria which have been used in earlier studiesto distinguishamongcells types in cat visual cortex are the following: I) Simple cells have one or more spatially distinct activating regions which can be plotted with stationary flashes. Complex cells do not show such separation. 2) The responseswithin eacharea summate with increasingstimuluslength in simplecellsbut not in complex. 3) Low spontaneousactivity is a property of simplecells. 4) Simple cellsare most numerous in layer 4, while complex cells are found primarily above and below this layer. 5) Simplecells have relatively smallreceptive fields in comparison with complex cells. 6) Hypercomplex cells form a separateclassdistinguished from others by virtue of the fact that they are strongly stopped for stimulus length along the axis of orientation. The distribution of cell properties and the histological data make evident the following points: I) The measure of subfield overlap (Fig. 20) shows a bimodal distribution suggesting two
I
1317
classesof cells. Only 36 cells appear in an area commonto both classes.2) Summationfor stimulus length, as shown in Fig. 17, doesnot differentiate S-type from CX-type cells. 3) The distribution of cells according to spontaneousactivity is unimodal. Cells classifiedas Stype and CX type show a considerable overlap so that this measurediscriminatesthese two classesonly in layers 5 and6.4) There isextensive intermingling of S-type and CX-type cells in almostall layers of cortex. 5) Receptive-field width shows a broad distribution, which appearsto be unimodal. CXtype cells have, on the whole, larger overall receptive fields than S-type cells, but there is an overlap of field size for a considerableportion of the two populations. A better differentiation between these classesis obtained, however, when subfield size to one sign of contrast change is compared. These observations suggestthat with the measuresdetailed, S-type and CX-type cellscan best be distinguishedon the basisof the spatialseparation of the light and dark responseareas. This measure does not, however, establish unique classes. Our qualitative observations confirm that Stype cells have inhibitory sidebandswhile CXtype cells lack this property. S-type cells with more than one excitatory subfield seemto have inhibitory sidebandsfor each of their subfields. Since mapping of these regions is difficult, it cannot be usedasan efficient meansof classifying cells into the S and CX categories. We have not studiedthe inhibitory property in sufficient detail to make quantitative and distributional statements about it. The major problem we encountered in our classification was with the third memberof the hierarchy as originally proposed by Hubel and Wiesel (9): the hypercomplex cell. While it is clear that examples of cells with hypercomplex properties can be found in area 17in the monkey, the strength of end stopping does not seemto provide a satisfactory criterion for classifying cellsin the monkey. The extent to which one can observe inhibition along the axis of orientation forms a unimodal, skewed distribution. Many, although not all, of the moderately and even strongly stoppedcellscould be clearly classifiedas S or CX type. End stopping, however, is not evenly distributed in the cortical layers. Significantly more cellsare stoppedin layers l-3 than in layers 5-6. This is an agreementwith the reports by Hubel and Wiesel (10) and Poggio(15). On the basisof this one might suggestthat the hypercomplex attribute is the product of a general inhibitory network, the effectiveness of which diminishesfor progressively deeper cells in cortical gray matter.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
nations of subfieldsin S-type cells which do not appear. For instance, cells in our samplewhich have two or more spatially separate fields responding to the same sign of contrast are extremely rare or nonexistent, although it is true that such combinationsmight be difficult to recognize. In spite of this, it is fair to say that some combinations have a high probability of occurrence while others have a low probability.
CORTEX
1318
SCHILLER,
FINLAY,
VOLMAN
a specific analysis of the visual input and 2) that they selectively project to other brain sites, one could perhaps better understand their function. In the subsequent papers we will attempt to contribute to the first question. Assessment of the second awaits further work. The major findings in this study and the inferences drawn from them may be summarized as follows: a) Several different types of S-type cells may be discerned in monkey striate cortex. The subfields of these cells appear to be produced by excitatory action. b) A hierarchy may exist among S-type cells. c) S-type cells are predominantly unidirectional, which suggests that the cortical transformation for directionality may occur at an early stage in striate cortex. d) S-type units, including the single contrast, S, cells are found in all cortical layers. c’) The only measures used in this study which yield well-separated distributions for S-type and CX-type cells are those which specify the spatial organization of the light- and dark-response areas of the receptive field. f) Unidirectional and bidirectional CX-type cells represent two distinct classes. g) Tonic cells, which are excited by only one sign of contrast change, form a third class of cells. h) The measure of the inhibition along the axis of orientation forms a continuous distribution which, therefore, does not yield a separate class of hypercomplex cells in the monkey’s striate cortex. ACKNOWLEDGMENTS
The authors thank Andres Polit, Cynthia Richmond, Kathy Anderson, and Louis Porter for their help on this project. This research was supported in part by grants from National Institutes of Health (EY00676) and the Alfred P. Sloan Foundation (72-4-l).
REFERENCES 1.
2.
3. 4.
5.
BISHOP, P. 0, COOMBS, J. S., AND HENRY, G. H. Responses to visual contours: spatiotemporal aspects of excitation in the receptive fields of simple striate neurones. J. ph~~iol., London 219: 625-657, I97 I. BISHOP, P. O., CUOMBS, J. S., AND HENRY, G. H. Interaction effects of visual contours on the discharge frequency of simple striate neurones. J. Physiol., London 2 19: 659-687, 197 1. BISHOP, P. 0. AND HENRY, G. H. Striate neurons: receptive field concepts. Invest. Ophthulmol. 11: 346-354, 1972. BROOKS, B. AND JUNG, R. Neurophysiology of the visual cortex. In: Handbook of Sensory Ne rrrop h ys iology , edited by R. Jung. Berlin: Springer, 1973, vol. VII/3B, p. 325-440. Dow, B. M. Functional classes of cells and their laminar distribution in monkey visual cortex. J.
Ncurophysiol.
37: 927-946,
1974.
6.
HARIING, J. K., BADE, B. J., AND ATENCIO, E. W. The organization of retino-geniculate and geniculostriate pathways in the grey squirrel (Sci-
urus curolinesis).
7.
Am.
Assoc.
Anat.,
87th,
1974.
D. H. AND WIESEL, T. N. Receptive fields of single neurones in the cat’s striate cortex. J. Physiol., London 148: 574-591, 1959. 8. HUBEL, D. H. AND WIESEL, T. N. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex. J. Physiol., London 160: 106-154, 1962. 9. HUBEL, D. H. AND WIESEL, T. N. Receptive fields and functional architecture in two nonstriate visual areas ( 18 and 19) of the cat. J. Neurophysiol 28: 229-289, 1965. 10. HUBEL, D. H. AND WIESEL, T. N. Receptive fields and functional architecture of monkey striate cortex. J. Physiol., London 195: 215-243, 1968. HUBEL,
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
The extent to which end stopping represents a transformation of the L,GN input to visual cortex is a knotty problem. It appears that the majority of cells in the LGN are strongly inhibited by their surrounds when long bars or edges are used, more so than the population of cortical cells we studied. On the basis of this one might consider the possibility that the significant cortical transformation produces a lack of end stopping rather than its opposite. This may be produced, as suggested by Hubel and Wiesel (8), as a result of a great deal of convergence of LGN cells on cortical cells. Strongly stopped cells in cortex may be produced either by a sparse input from the LGN or by a special inhibitory circuit in cortex. A more thorough study is needed to assess the nature of this transformation. Finally, we might pose the question of what the functional significance of the various cell types might be. Why, for example, are so many S-type cells comprised of bipartite or multipartite excitatory fields? Does this arrangement permit a specific form of analysis of visual information? Assuming a hierarchy, with S-type cells providing the input to CX-type cells, why should this property be promptly lost in CX-type cells? Perhaps S-type cells are involved in the analysis of spatial frequency or some other parameter and send this information to another structure for further analysis. On the other hand, it is also possible that the spatially separate subregions observed in these cells serve no specific function, and may be an outcome of a relatively sparse input from the slightly misaligned regions of the multilayered I, GN terminal fields. Further pooling of S-type cell outputs would eventually get rid of this defect. If it could be shown that I) S-type cells perform
AND
MONKEY 11.
12.
13.
15. 16.
HUBEL., D. H. AND WIESEL, T. N. Laminar and columnar distribution of geniculocortical fibers in the macaque monkey. J. Comp. Neural. 146: 421-450, 1972. HUBEL, D. H. AND WIESEL, T. N. Sequence regularity and geometry of orientation columns in monkey striate cortex. J. Comp. Neurol. 158: 267-294, 1974. HUBEL, D. H. AND WIESEL, T. N. Uniformity of monkey striate cortex: a parallel relationship between field size, scatter and magnification factor. J. Comp. Neurol. 158: 295-306, 1974. MAFFEI, L. AND FIORENTINI, A. The visual cortex as a spatial frequency analyser. Vision Res. 13: 1255-1267, 1973. POGGIO, G. F. Spatial properties of neurons in striate cortex of unanesthetized macaque monkey. Invest. Ophthalmol. 11: 368-377, 1972. SCHILLER, P. H., FINLAY, B. L., AND VOLMAN, S. F. Short-term response variability of monkey striate neurons. Brain Res. 105: 347-349, 1976.
CORTEX
I
1319
17.
SCHILLER, P. H. AND KOERNER, F. Discharge characteristics of single units in superior colliculus of the alert rhesus monkey. J. Neurophysiol. 34: 920-936, 1971. 18. SCHILLER, P. H., STRYKER, M., CYNADER, M., AND BERMAN, N. Response characteristics of single cells in the monkey superior colliculus following ablation or cooling of visual cortex. J. Neurophysiol. 37: 181-194, 1974. 19. STONE, J. Morphology and physiology of the geniculocortical synapse in the cat: the question of parallel input to the striate cortex. Invest. Ophthalmol. 11: 338-346, 1972. 20. WATKINS, D. W., WILSON, J. R., SHERMAN, S. M., AND BERKLEY, M. A. Receptive field organization: further differences between simple and complex cells. Assoc. Res. Vision Ophthalmol.,
Sarasota,
21.
Fla.,
1975.
M. L., MACNICHOL, E. F., AND H. G. Glass insulated platinum microelectrode. Science 132: 1309-1310, 1960. WOLBARSHT, WAGNER,
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
14.
STRIATE
This article has been cited by 56 other HighWire-hosted articles: http://jn.physiology.org/content/39/6/1288#cited-by Updated information and services including high resolution figures, can be found at: http://jn.physiology.org/content/39/6/1288.full Additional material and information about Journal of Neurophysiology can be found at: http://www.the-aps.org/publications/jn This information is current as of June 3, 2013.
Journal of Neurophysiology publishes original articles on the function of the nervous system. It is published 12 times a year (monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 1976 the American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at http://www.the-aps.org/.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
You might find this additional info useful...
JOURNAL. OFNEUROPHYSIOLOGY Vol. 39, No. 6, November 1976. Printed
in U.S.A.
PETER
H. SCHILL,ER,
Depurtment Cambridge,
SUMMARY
BARBARA
of Psychology, Massachusetts
AND
L,. FINLAY,
Massachusetts 02139
Institute
CONCLUSIONS
1. The properties of single cells in striate cortex of the rhesus monkey, representing the visual field 2”-5” from the fovea, were examined quantitatively with stationary and moving stimuli. Three distinct classes of cells were identified: S type, CX type, and T type. 2. S-type cells were defined as those oriented cells which to the optimal direction of movement in their receptive fields exhibited one or more spatially separate subfields within each of which a response was obtained to either a light or dark edge, but not to both. Several different types of S-cells were distinguished: u) S-type cells for which moving edges revealed a single excitatory area within which a response was elicited by either a light or a dark edge but not by both. Most of these cells were unidirectional. 6) S,type cells for which moving edges revealed two spatially separate response areas, one of which was excited by a light edge and the other by a dark edge. Both regions responded to the same direction of movement. c) &-type cells which had two response areas, one of which was excited by a stimulus moving in one direction (at right angles to the axis of orientation) and the other, of opposite contrast, which responded in the opposite direction. (i) S-type cells which to one direction of movement showed two spatially separate regions sensitive to a light and dark edge and which in the other direction of movement had only one responsive area (either light or dark). c) Cells which had multiple spatially separate subfields (S,., types). 3. CX-type cells were defined as those oriented cells which in their receptive fields exhibited no spatial separation for light- and dark-edge responses; they discharged to both edges in the same direction of movement and in Received 1288
for
publication
December
22, 1975.
AND
SUSAN
F. VOLMAN
of Technology,
the same spatial area. Flashing stimuli elicited both on and off responses throughout the receptive field. CX-type cells were predominantly of two types: those which were selective for direction of stimulus movement and those which were not. 4. A third class of cells (T-type) were those which were excited by only one sign of contrast change and responded in a sustained fashion even when there was no contour within the receptive field. These cells were poorly or not at all oriented; some of them were selective to wavelength. 5. Quantitative comparisons showed the following differences between S-type and CX-type cells: u) S-type cells had smaller receptive fields than CX-type cells but the populations overlapped considerably. Receptive-field size was smallest in layer 4c. In all other layers S-type cells had the same size fields. CX-type cells, by contrast, tended to have larger fields in layer S-6 than 2-3. b) The spatial separation between light and dark response areas was the best criterion for distinguishing S-type and CX-type cells. The distribution of this measure disclosed two populations of cells with relatively limited overlap. c) In layers 2 and 3, both S-type and CX-type cells had low spontaneous activity. In layers 5 and 6, S-type cells retained this attribute. Most CXtype cells, however, showed significantly higher spontaneous activity in these bottom layers. Thus the criterion of spontaneous activity discriminated S-type and CX-type cells in layers 5 and 6 but failed to do so in the upper layers. cl) S-type cells were found in all cortical layers. CX-type cells were absent in layer 4c. 6. The measure of inhibition along the axis of orientation, typically used to classify cells into the “hypercomplex” category, did not establish a distinct subcategory of cells. This property was observed in both S-type and CX-type cells
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
Quantitative Studies of Single-Cell Properties in Monkey Striate Cortex. I. Spatiotemporal Organization of Receptive Fields
MONKEY
STRIATE
INTRODUCTION
Following the pioneering work of Hubel and Wiesel (7, 8), striate cortex has been subjected to extensive analysis with single-unit recording techniques (l-5, 1 l-14). The input to this structure from the lateral geniculate nucleus (LGN) is transformed in five notable ways. The first of these produces specificity for the orientation of contours within the receptive field. The second transformation is that of direction selectivity. While in the LGN of the cat and monkey cells respond to stimuli moving in any direction, in cortex many cells are responsive to movement only in one direction. The third transformation involves the combination of responses to contrasts of opposite sign. Most LGN cells in the center of their receptive field are excited by either light increment or light decrement, but not by both. In cortex these attributes are combined so that many cells respond to both kinds of stimulus change within a single region. The fourth transformation unites the input from the two eyes, resulting in cells which can be binocularly activated. The fifth transformation alters the spatial frequency response of cells: in the striate cortex there is a sharply increased selectivity for spatial frequency when compared to what is found in the LGN (14). All five transformations may be seen in a single cell. In this series of papers we will present the results of experiments in which we analyzed the visual cortex of the monkey in terms of these transformations. In the first paper we will examine the response properties of single neurons in striate cortex *to direction of movement, contrast, and stimulus length, with special emphasis on simple cells. The second paper will present data regarding orientation selectivity and ocular dominance. The third paper will be concerned with selectivity for spatial frequency. The fourth paper will examine the properties of the corticotectal cells in area 17. In the last paper we will consider the relationship among the different measures we used and how these m easure s reflec t on various models of visual corte X. Work on the cat has disc losed several cri teria whe rcby simple and compl ex cells can be distinguis hed (1, 8, 20). Simple cells have relatively
I
1289
small receptive fields. These fields can be plotted with stationary flashing spots of light; such a plot discloses spatially separate activating regions for the onset (on-response) and the turning off (off-response) of a flashing light. The response within a given region summates as the size of the stimulus increases. Moving stimuli may demonstrate either one such activating region or several spatially separate activating regions. Furthermore, simple cells have low spontaneous activity and show a preference for relatively slow velocities of stimulus movement. In the cat, cells with simple receptive fields are most numerous in layers 3, 4, and 6 of striate cortex (8). Complex cells have larger fields and respond more poorly to flashes. The activating region shows no subdivision into spatially separate zones: when a cell responds to them, flashing stimuli elicit both on- and off-responses throughout the field; similarly, a response is obtained to both the light edge and the dark edge of moving stimuli within this one region. These cells are more often binocular, tend to have greater spontaneous activity, especially in cortical layers 5 and 6, and respond over a greater range of velocities. In contrast to work on the cat, simple cells in the monkey have not been studied in detail, primarily because they have been encountered less frequently. Hubel and Wiesel (10) report that only about 8% (25 of 272) of the cells they have studied in the monkey fell into this category. Poggio (15) classified 7% (12 of 224) and Dow (5) 9% (21 of 234) of the cells as simple. Considerable uncertainty exists regarding the location of these cells in monkey striate cortex. Hubel and Wiesel (10) found that of the 25 cells they classified as simple, 23 were in layer 4. By contrast, Poggio (15) reports only 1 of 12 in this layer. Dow (5) shows a relatively even distribution of simple cells among the cell layers. According to Hubel and Wiesel (IO), complex cells in the monkey are least common in layers 4b and 4c. Poggio’s data show them equally distributed. Subsequent to their initial discoveries, a third class of cells has been identified by Hubel and Wiesel (9) in the cat; these cells were called hypercomplex, and it was suggested that they were formed by excitatory and inhibitory inputs from complex cells. The major criterion for placing the hypercomplex cells into a distinct class was the observation that their responses are inhibited by stimuli extended along the axis of orientation, thereby rendering them selective for stimulus length, an attribute absent in simple and complex cells. Several qualifications have been made subsequently regarding this cell type (3). Further investigation disclosed several other
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
and was continuously distributed. However, this inhibition was more prevalent in the superficial layers than in the deep ones. 7. The results are suggestive of a hierarchy among S-type cells. Lowestin this hierarchy are those S-type cells which could be activated by only one kind of moving edge. From these, S-type cells with two or more subfields may be constructed.
CORTEX
1290
SCHILLER,
FINLAY,
VOLMAN
which within their activating region showed no spatial separation; the areas which responded to light increment and light decrement overlapped. Specificity for direction in these cells was similar for moving light edges (light increment) and dark edges (light decrement). In what follows we present qualitative and quantitative data describing the properties of various types of cells found in monkey striate cortex as determined by moving edges and stationary, flashing stimuli. We will place special emphasis on S-type cells as here defined. Distributions of cells according to a variety of response measures will be used to scrutinize the validity of classificatory schemes. Finally, these data will be considered in terms of the anatomical distribution of cell types within the layers of striate cortex. METHODS
Single-unit recording procedures A total of 42 rhesus monkeys (Macaca mulatta) were used to obtain the data reported in this paper. All 42 were studied while the animals were anesthetized and paralyzed. In addition, two of these animals had also been recorded from while alert, with one eye surgically immobilized. ANESTHETIZED ANIMALS. Surgical procedures were carried out under Pentothal anesthesia. After tracheotomy and vein cannulation, animals were placed in a Kopf stereotaxic apparatus with raised eye and ear bars. In 32 animals a closed chamber of 10 or 16 mm diameter was implanted over area 17 in a region where receptive fields were found 2”-5” from the fovea in the lower visual field. The dura mater overlying the visual cortex was removed in approximately half the animals and was left intact in the others. We found that successive penetrations can be undertaken with greater ease, rapidity, and with fewer episodes of electrode breakage when dura is removed. This procedure also permits the use of finer electrodes. On the other hand, recordings are not quite as stable and cortex deteriorates more rapidly. In 32 of the animals penetrations were made perpendicular to the surface of the brain. In 10 monkeys tangential penetrations were made which were between 10” and 25” to the surface of the brain, in a manner similar to that described by Hubel and Wiesel (12). Animals were infused with Flaxedil (40 mg/h) dissolved in 5% dextrose in Ringer and were artificially respired. End-tidal CO, was monitored and maintained at 4.&4.5%. Pentothal anesthesia was discontinued after the first 2-10 h
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
classes of cells in striate cortex. Prominent among these are unoriented cells, which may be further divisible into several subgroups (5, 10, 15). One of the problems with these classifications is that few of the dimensions which have been used for assigning cells to categories have been studied quantitatively. Thus it is not known, especially in the monkey, to what extent the measurements used form distributions which justify the classificatory schemes. The assumption of classes can be inferred quantitatively when the measure on the basis of which the classification is made yields separate groupings as demonstrated, for example, by a bimodal distribution. If, however, a distribution is continuous, one cannot, of course, claim two distinct classes. An example in point is binocularity. Since the extent to which cortical cells are binocular forms a relatively continuous distribution, this property has been believed not to reflect two distinct classes of monocular and binocular cells. Thus, issues about binocularity are typically formulated in statistical terms asking, for instance, how the ocular dominance distribution is affected by such variables as brain site or environmental manipulation. The question may be raised: to what extent will the quantitative measures of single-cell receptive-field properties disclose distributions which justify the classificatory schemes hitherto proposed? In particular, we will examine distributions based on such measures as subfield separation, directionality, contrast independence, and spontaneous activity. In addition to the problem of classification, the criteria which have been used to group cells in striate cortex into the simple, complex, and hypercomplex classes have been under some debate (3, 11). It appears that a portion of the cells considered as simple by some investigators may not have been so designated by Hubel and Wiesel (8) who were the first to describe and name these cells. Almost unavoidably, perhaps, each investigating group has developed its own set of criteria. Since we may too have done so, we decided to call cells in the monkey which we believed to be simple, S-type cells. Those which we thought analogous to complex cells we called CX-type cells. We found that the most reliable classification was achieved on the basis of the spatial location of responses to light increment and light decrement. For this reason we defined our S-type cells, as did Hubel and Wiesel (8), as those orientation-selective cells whose receptive fields had one or more distinct subfields within each of which a flashing stimulus or a stimulus moving in one direction elicited a response to either the increment or decrement of light, but not to both. CX-type cells were defined as those
AND
MONKEY
STRIATE
In two animals one eye was MONKEYS. surgically immobilized by transection of the 3rd, 4th, and 6th cranial nerves. Skull screws and recording wells were implanted prior to recording as previously described (17). During recording sessions the head was restrained, permitting the study of cortical receptive fields via the immobilized eye. The prime purpose of this study was to determine whether or not the receptivefield properties of single cells in visual cortex of the alert animal in our situation were comparable to those of the paralyzed, anesthetized animals. ALERT
Stimulus delivery sys terns Two separate stimulus delivery systems were used, one manual and one computer operated. MANUAL SYSTEM. The manual system consisted of an optic bench, a control box, and assorted power supplies. The optic bench had the following elements: a) tungsten ribbon filament lamp; b) condenser;c) motor-driven aperture permitting adjustment of stimulus length, width, and orientation; d) projection lens; e) X-Y mirror galvanometer. The image was projected to a tangent screen 57.3 inches from the animal. The control box contained a joy stick, which operated the X-Y galvanometers, and several switches which activated the aperture motors. The manual system was used to localize the receptive field on the tangent screen and to determine qualitatively the basic attributes of receptive fields such as orientation, binocularity, directionality, and the location of subfieldsin the S-type cells.
I
1291
bench were driven with stepping motors activated either manuallyor by a PDP- 1l/20 computer. The aperture in this case consistedof two sets of leaves driven by two stepping motors which could produce rectangularstimuli. Stimuli could be made to increase or decreasein size either symmetrically with respect to the center of the optical axis or asymmetrically by keeping one leaf fixed at the center of the optical axis. The next element on the optic bench was the projection lens which wasfollowed by a shutter. After the shutter a right-angle prism permitted the movement of the stimulususingeither a galvanometer or a steppingmotor. The galvanometer was used in conjunction with a waveform generator for smooth stimulusmovement on the tangent screen. The steppingmotor was usedto place stimuli into various fixed positions on the screen. Following this was a dove prism which was rotated with a stepping motor to produce changesin stimulusorientation and direction of motion. All the motor-driven elementsin this system could be manipulated either manually or by computer. The PDP- 11was used and was programmedto allow the manipulation of each of several stimulus dimensions such as stimulus orientation, stimuluslength, and stimuluswidth. One of the major features of the computer program we used, which was developed by A. Polit, wasthat it enabledusto presentstimulus conditions in randomizedorder. The determination of orientation specificity, for example, was accomplished by having a bar or edge move acrossthe receptive field in a numberof different orientations, all of which were presentedautomatically in a randomorder after specificationof the rangeand number of trials. For someof our quantitative measuresthis feature was deemed essentialdue to the inherent responsevariability of single cells in visual cortex. Our data (16) show that the average short-term variability of cortical neuronsto repeated presentationsof an optimal stimulusis 35.4 using the formula: standard deviation/mean x 100. On collecting the data for a given run, the results were displayed on a Tektronix 611 storage oscilloscope; total, mean, and standarddeviation of the numberof spikesobtained to each stimulus presentation could be displayed on command. Thesedata could also be printed out on the Teletype or stored on DEC tape. The stimuli used were 0.8-l .6 log units above cell response threshold. Background illumination was 0.8- 1.2 cd/m?
SYSTEM. Stimulusde- PROCEDURE livery for detailed analysiswas accomplishedby Each unit was first studied by manualoperaa second optic bench. The elements on this tion of the first optic bench. The response COMPUTER-CONTROLLED
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
and subsequently animals were maintained on a 70%-30% mixture of N20 and OZ. Following application of hyoscine, contact lenses of proper curvature were applied to bring the eye in focus on a tangent screen 57.3 inches away. Using a reversible ophthalmoscope the fovea of each eye was projected onto the screen. In most experiments 3-mm artificial pupils were placed in front of the eyes. Animals were typically studied over a period of 2 days. During the night contact lenses were removed for a few hours to reduce cornea1 clouding. Even so, in some animals slight opacity was observed by the end of the 2nd day of recording. When this occurred the cornea1 epithelium was removed by scraping, which resulted in clear optic media (18). Single cells were recorded using glass-coated platinum-iridium microelectrodes (2 1).
CORTEX
SCHILLER,
1292
characteristics quantitatively rangement.
FINLAY,
were subsequently assessed with the computer-driven ar-
Manual assessment
Computer-driven
assessment
The following procedures were used to assess the response characteristics of single cells quantitatively. Each step described below was carried out as a single uninterrupted sequence, initiated by the necessary commands via a Teletype* ORIENTATION. Most commonly we began the detailed study of a cell by initiating a computer command which resulted in having a bar or edge sweep across the receptive field in different orientations in a randomized sequence. The results obtained in this fashion will be described in detail in the second paper. On completion of an orientation sequence the data were stored and displayed on a 611 storage oscilloscope enabling us to see the optimal orientation and the sharpness of orientation tuning. STIMULUS LENGTH. For this series the stimulus was set to the orientation which provided the best response during the orientation sequence described above. The length of the stimulus was varied in a random order, typically over eight steps using stimuli of fixed widths and varying lengths ofO.1, 0.2, 0.4, 0.8, 1.6, 3.2, 6.4”, with 10 or 20 measures in all at each step. Included in this series was a “no stimulus” condition used to obtain a measure of the response rate in the absence of a visual stimulus. MOVING EDGES. To assess the response of units to light (light increment) and dark (light decrement) edges, to determine the spatial location of the response areas, and to assess degree of directionality, single edges or rectangles were swept across the receptive field. Commonly 1” by 1” stimuli were used and most units were tested with both a light square on a dark background and a dark square on a light background.
VOLMAN
These stimuli were moved back and forth across the receptive field in the optimal orientation, usually over a range of 3” at a velocity of 2”/s, for 30 repeated trials. For units with large fields or complicated responses, single edges were used and for strongly “stopped” cells, stimuli with lengths shorter than 1” were introduced. During data collection the responses were displayed as they occurred on the 611 storage oscilloscope in the form of a raster, with each response generating one dot and successive stimulus sweeps appearing on successive lines of the oscilloscope. At the termination of the sequence the data were stored and displayed in the form of a cumulative-response histogram; 256 bins were used, which for 3” back-and-forth movement at 2”/s meant 11.7 ms/bin or 0.023’ spatial resolution per bin. FLASH ES. The responses of cells to stationary flashes were studied using two procedures. One was a single sequence of 30 repeated flashes of an optimal stimulus in the best location with a 1.5 s on and a 1.5 s off cycle. For the second method, several spatially separate sites were flashed. The stimulus was typically a O.09”-wide bar in the optimal orientation of the receptive field. Stimuli were flashed over a range of either 0.8” or 1.6, depending on receptivefield size, using eight different equally spaced positions; 20 trials, with a 1 s on, 1 s off cycle, were obtained for each position in random order. STATIONARY
RESULTS
The results reported in this section are divided into five parts. In the first we will describe the response of various kinds of cells in monkey striate cortex to stationary and moving edges. The second section will deal with the size and shape of receptive fields. In the third section the distributions of cells according to a variety of quantitative measures will be presented. In the fourth section the relationship between unit responses and the site of recording in the cortical layers will be analyzed. The last section briefly describes results obtained in alert animals.
Examples of cell properties assessedby responses to stationary and moving stimuli We defined S-type cells as those oriented cells whose receptive fields had one or more spatially distinct regions within each of which a movingedge or a flashing stimulus elicited a response to either light increment or light decrement. By contrast, CX-type cells were defined as those oriented cells which to either moving edges or stationary flashes of light responded to both light increment and light decrement throughout their
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
When a single cell was clearly isolated and properly triggered by the Schmitt trigger, we began by determining: I) the location and size of the receptive field; 2) the dominant eye; 3) the best orientation; 4) the optimal direction of movement; 5) the responsiveness to light and dark edges or bars; 6) the effect of different stimulus lengths and widths; and in some cases, 7) the response to stimulus velocity within a range of O.S-50%; and 8) the response to colors using red, green, blue, and yellow filters. Some of these measures were again reassessed manually following the computer-operated sequence.
AND
MONKEY
STRIATE
When a single photocell respondstransiently to any sudden change in illumination,
a light or
dark rectangle sweepingacross it will generate four pulsesin a histogram(Fig. IA), two in each direction of movement, one for the light edge and the other for the dark edge. If the photocell circuit is arranged to produce a pulse to light increment alone, only those peaks will be seen which
are labeled
L. If, furthermore,
I
we have
a
FIG. 1. Photocell responses to the light and dark edges of moving rectangles. A: light rectangle on dark background, single photocell responding to both light and dark edges. B: light rectangle on dark background, two spatially displaced photocells, one responding to the light edge and the other to the dark edge. C: dark rectangle on light background, photocell arrangement same as B. Direction of movement across photocells is shown by arrows; turnaround point is indicated by triangle. L, response to light edge; D, response to dark edge. Time is left to right throughout. To facilitate comparisons, thin vertical lines are drawn for each edge centered on responses appearing in the first histogram.
I
1293
the photocell respond to only one direction of movement, only one peak would be seen. Not so obvious is what happens when we place two photocells sideby sidewith somespatial separation between them and have the first respond transiently only to light increment and the second only to light decrement. The responses elicited by moving white and black rectangles traversing the two photocells are shown
in Fig. 1B and C. As can be seenthe pulsesare now asymmetrically displaced with opposite asymmetriesfor the light and dark squares. If responselatency and the total distancetraversed by the moving stimuli are known, the relative location of the responseareascan be calculated on the basisof the temporal separationbetween the pulses(1). CELLS. Our findings suggestthat S-type cells with a variety of responsecharacteristics can be distinguishedin area 17 of the monkey. From the nature of their responsesseveral inferencesmay be maderegardingthe organization of visual cortex. Qualitatively we could discern sevendistinct cell types from our total sampleof 245 S-type cells. An example of each cell type will be shown. Figure 2 showsthe simplestresponsewe obtained to moving stimuli from S-type cells. This cell dischargedonly to a dark edgetraversing the field and did so only in responseto one direction of movement. The stimuluswas moved at right anglesto the axis of orientation so asto elicit the optima1response.Figure 2,1 showsthe response to a white, lo square. In Fig. 2,2, stimuluscontrast is reversed; a black 1” square is swept across the receptive field. In both cases responseoccurs to light decrement, that is, to the dark edgeonly. In the drawing of this figure and subsequentones, the original 256 bins in the computer display were reduced to 128by combining adjacent pairs of bins. In Fig. 2,3 the responseproperty of this cell to moving stimuli is drawn in schematicmanner. It showsthe approximate size of the field and the kind of responseelicited. The arrow indicates that the responsewas unidirectional. This unit may be describedthen asa unidirectional S-type cell, sensitive to a dark edge. This cell respondedvigorously to the turning off of a thin rectangular light bar flashingin the center of the field, as shownin Fig. 3,1. Flashingthis stimulus anywhere but in the center of the field produced no responseat all. A large rectanglecentered on the field, when flashed, also failed to elicit any S-TYPE
response.
However,
when
the center-activat-
ing region was continuously illuminated, the superimposedflashing of a large rectangle did evoke a responseto its onset. These observa-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
receptive fields. For these cells, direction selectivity was not affected by the contrast of the moving stimulus. The results described in this section were obtained by I) sweeping light and dark rectangles or single edges across the receptive field, and 2) flashing stationary targets into various parts of this field. The use of both light figures on a dark background and dark figures on a light background was considered an important reciprocal control for moving stimuli, and the data shown always include examples of both. In most cells a close correspondence was found between receptive-field maps obtained with small stationary, flashing spots and with moving stimuli. As an introduction to the data obtained with moving rectangles, the first figure describes what such responses would look like when the stimuli are swept across photocells attached to the tangent screen where the receptive fields of cells were to be studied.
CORTEX
1294 UNIT
SCHILLER,
FINLAY,
AND VOLMAN 100
78-4-12.30 I
UNIT
-
78-4-12.30
I
W
ON
ON I DEGREES
2 OF
4
3 VISUAL
ANGLE
2. Unidirectional S-type cell located 3” from the center of the fovea which, with moving stimuli, responds only to a dark edge. 1: response to a 1” light square. 2: response to a 1” dark square. Arrows indicate direction of movement across receptive field. Extent of stimulus movement: 3” in each direction, turnaround point indicated by triangle. Velocity: 2”/s. Histogram time is left to right throughout. Vertical lines indicate position on the histogram of the response that a single photocell would give to each edge. The position of these lines has been adjusted for unit response latency. The total number of bins across equals 128, reduced from the original 256 by adding adjacent bins. Number of trials per histogram, 30. L, light-edge response; D, dark-edge response; sp, spikes. 3: schematic drawing showing size and approximate spatiotemporal response properties of the receptive field. Arrow indicates direction of movement across field: the upper part of this figure depicts response to movement in one direction and the lower part response to movement in the opposite direction. FIG.
FIG. 3. Flash response of unit in Fig. 2. Top histogram shows response to a thin bar flashed in the center of the receptive field. Bottom histogram shows response to a large rectangle which was flashed while the center was continuously illuminated. Flashing of this same large rectangle without the small rectangle tonically illuminating the center did not elicit a single response in 30 trials.
activating region of the field. Cells of this sort are commonin area 17of the monkey, and their counterpart, cells responding to a light edge only, are equally numerous. Of the 245 S-type cells studied, 67 were of this type; 34 had light fields and 33 had dark fields. Over 80% of them were unidirectional. We called these S, or single-contrastcells. Figure 4 showsthe responsepropertiesof another cell. It respondedto both a light edgeanda dark edge. The responseswere elicited from two different regions in the field, as is evident from the different locations of the peaks in Fig. 4,l and 4,2. Had they been in the sameplace, the tions suggestthat the cell has a surround which responseswould have centered on the vertical is opposite in sign to the center. This surround lineswhich showwhere a photocelllocatedin the region, however, is unresponsive to moving center of the field would have respondedto each stimuli and to small stationary flashing spots. edge of the stimulus (Fig. IA). This cell is similar to LGN cells in that it hasan Figure 4,3 shows the schematicorganization antagonistic center-surround organization; the of this cell to a moving stimulus, demonstrating surround can be potentiated by illuminating the spatially separate light- and dark-activating recenter region. It is also similar to LGN units in gions. Both of these regions appear to be exthat in its central activating region it responds citatory sincecontrast reversal doesnot alter the only to one sign of contrast. However, in con- responseto each edge. Only the temporal setrast to units in the LGN, this S-type cell is quenceof events changesbecauseof the way the orientation and direction selective, has low stimulusedgesof oppositecontrast move across spontaneousactivity, and dischargesto moving the two subregions(see Fig. 1). The centers of stimuli only while a dark edgemoves acrossthe the two regionsare 0.29” apart. When stationary
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
m
MONKEY
STRIATE
CORTEX
I
1295
UNIT 91-l-22.27
UNIT 72-8-13.55
4 , 0B
OF
VISUAL
z
3
4
5
6
7
9 dep
ANGLE
FIG. 4. Unidirectional &type cell with separate subfields for light- and dark-edge responses. 1: response to a 1” light square. 2: response to a 1” dark square. 3: schematic drawing of receptive field showing size. separation, and directionality of the two subfields. The two subfields responding to opposite contrast respond to the same direction of movement (all parameters as in Fig. 2).
flashes are used a similar spatial separation of the light region (on-response) and dark region (off-response) is revealed. The cell is essentially unidirectional. We called these & or double-field cells; 47 units in our sample had this property. The cell in Fig. 5 shows another commonly occurring property in the cell population of area 17 of the monkey. This cell responds to both light and dark edges, but the direction of the response is affected by contrast. Both the light and the dark edge responses are unidirectional, but they are specific for opposite directions of motion. The locations of these responses are spatially separate, as shown in the schematic spatiotemporal drawing in Fig. 53, where the upper part of the figure shows response to movement in one direction and the lower part, response to movement in the other direction. The spatial separation between the two regions can be derived two ways. It can, first of all, be determined by hand plotting using slowly moving edges or small stationary flashing spots. Second, it can be calculated. This calculation, however, must take into account the response latency of the discharge. This we typically obtained by measuring response latency to flashing
FIG. 5. &-type cell demonstrating interaction between contrast and direction. 1 and 2: responses to I” light and dark squares. 3: schematic drawing of spatiotemporal response. The two subfields responding to opposite contrast also respond to opposite directions of movement.
stimuli of equal intensity. In some cells latency was calculated from responses obtained to different stimulus velocities, as described by Bishop, Coombs, and Henry (1). For this cell, then, the light region is activated by one direction of movement and the dark region by the other. Reversing the sign of contrast between the stimulus-background does not alter this relationship. We called these Z$ or interaction cells because they demonstrate an interaction between direction and contrast. Such units may be thought of as responding best to a single edge moving back and forth across the field which, for this cell, has the light region on the left and the dark region on the right side. When the contrast is reversed, moving the edge within the field elicits no response. When tested in this fashion this is exactly what happens; 36 cells in our sample had this kind of interaction between contrast and direction. Figure 6 shows another common S-type cell. It is one which discharges to a dark edge in both
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
I DEGREES
1296
SCHILLER,
UNIT 84-7-9.15
L-3
FINLAY,
fields. The cells shown in Figs. 4, 5, and 6 have similar spatial arrangements.The directionality of the responseamongthesecells is quite different, however, and is not predictable from the maps obtained with the stationary flashing stimuli. The schematic diagram showsthe spatial arrangement and relative responseamplitudes of the regions of this field. The light area is unidirectional. The dark area may be thought of as a singlebidirectional subregionor as two subregions, one respondingin one direction and the other responding in the opposite direction. Complementary cells which have bidirectional light and unidirectional dark areas are also common. We have a total of 31 suchasymmetric cells in our sample, 14bidirectional for light and 17bidirectional for dark. Thesewere called S4or U-B cells, where U standsfor unidirectionality for one contrast and B for bidirectionality for the other contrast. Figure 7 shows a cell which produces a bidirectional response for both light and dark edges. Both of the light areas and both of the dark areasappear to be in the samelocation for the two opposite directions of movement; 2 1 UNIT 92-I-27.15
FIG. 6. &-type cell with unidirectional light edge and bidirectional dark-edge responses. I and 2: responses to light and dark 1” squares. 3: response to a 0.09’ bar flashed in different locations. Open circles are response to onset of light; closed circles show off-response. 4: schematic drawing of receptive field.
directions but to the light edgein only one direction. This occurred in the samemanner whether 3 one moved a light squareon a dark background (6,l) or a dark squareon a light background(6,2) across the field. Figure 6,3 shows the results of flashing a stationary 0.09” light bar in various parts of the receptive field. The light bar was flashed in six different locations 20 times each in random order. The meannumber of responsesare plotted for the first 250msafter the light is turned on and off. These data clearly show the separatespatial FIG. 7. &-type cell with bidirectional light and dark locations of the light (on) and dark (off) response resnonses.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
3
AND VOLMAN
MONKEY
STRIATE
1297
stimuli they could be mistaken for CX-type cells. Using single edges again, Fig. 9 shows a cell in area 17 with a bidirectional response, which, in both directions, shows three activating regions; a central dark edge region flanked by light regions. The first subfield responds well to leftright movement; in the opposite direction this region is barely observable; 17 cells in our sample had multifield properties of this sort, and were called S, or flanked-field cells. The subfields of some of these cells could not be activated with both directions of movement. Thus some cells had a tripartite field in one direction and only a bipartite field in the opposite direction. The cells we have considered up to now had relatively distinct properties. It would be misleading, however, to claim that all S-type cells were as well defined as these. We have found a minority of S-type cells with intermediate characteristics, suggesting that the various Stype cells, inasmuch as they fall into groups, do not form unique categories. Two examples of such cells appear in Fig. 10. The first (Fig. 1OA)
UNIT 59-3-17.85
FIG. 8. !$type cell with four subfields. 1 and 2: responses to moving single edges. 3: schematic drawing showing that light and dark edge subfields are in different locations for opposite directions of movement.
I
b
FIG. 9. Bidirectional, multiple subfield S-type cell. 1 and 2: responses to moving single edges. 3: schematic drawing of receptive field.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
S-type cells had this characteristic with some variability in the location of the fields as a function of direction of movement, and were called &, or B-B cells. The cells shown so far are relatively common in area 17 of the monkey. Figure 8 shows a less common cell type with rather interesting properties. Because of the complexities of the response and the size of the receptive field, the data shown were obtained by moving only a single edge across the field, with the contrast of the light and dark regions reversed for Fig. 8,l and 8,2. The field was also carefully hand plotted to verify the relationship between the light- and dark-activating regions. The schematic drawing in Fig. 8,3 shows two sets of light and dark regions which are located in different places depending on the direction of movement. Five cells of this type were observed in our sample, and were called Ss or double interaction cells because of the two regions of interaction between direction and contrast. Most cells with this kind of spatial arrangement produce both on- and off-responses to flashing spots throughout their receptive fields. For this reason, if they were mapped only with stationary
CORTEX
SCHILLER,
1298
FINLAY,
AND
VOLMAN
A UNIT 91-l-27.50
FIG.
10.
TWO
L
A
0
L
examples of S-type cells with properties which fell in between some of those previously shown.
is one which responds predominantly to one edge; additional fields are evident, however, placing it in between a single-field S-type cell (Fig. 2) and an S-type cell (Fig. 7). In Fig. 1OB an example is given of a cell whose properties fall between an interaction (S,) cell and a cell responsive to both light and dark edges in both directions (S,). For none of the S-type cells we studied was it possible to predict the directionality of the response reliably on the basis of the receptive-field map obtained with stationary flashing spots. Cells which responded to both CX-TYPE CELLS. light and dark edges throughout their activating region and showed no interaction between direction of movement and stimulus contrast were classified as CX type. There were 342 CX-type cells in our total sample of 1,125 units. Figure 11 shows a unidirectional CX-type cell. It has a relatively large receptive field as shown both with moving edges and with flashing
stationary targets (Fig. 11,3). The data and the schematic drawing demonstrate overlapping light and dark regions. Figure 12 shows a similar cell although with low spontaneous activity and a smaller field. This unit is bidirectional with superimposed light and dark regions in both directions. The two cells in Figs. 11 and 12 conform to the typical complex cells so far described. The majority of CX-type cells are of this sort. We have found, however, that a number of these cells had rather small fields. Such a cell is shown in Fig. 13. By our criteria this is a CX-type cell, yet its field is no larger than that of S-type cells in the same region of the visual field. A small degree of interaction between contrast and direction, typical of &-type (interaction) cells, is also evident in this figure. INHIBITORY
RESPONSES
OF S-TYPE
AND
CX-TYPE
In the cells described so far the schematic drawings were derived from the prinCELLS.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
0
MONKEY
STRIATE
CORTEX
I
1299
UNIT 84-3-13.05
. I
, .2
. .3
. .4
. .s
. .6
.2
.?I
.4
.s
.6
FIG.
11.
. .7
6
.8
Unidirectional
cipal excitatory responses to increase or decreasein energy level produced by either moving edgesor stationary flashing stimuli. In this section we will deal with the inhibitory responsesin S-type and CX-type cellsas observed on the basis of a decrease in responsiveness, which presumably reflects inhibitory action. Bishop, Coombs, and Henry (2) have shown that the inhibitory side bands of simple cells in cat’s visual cortex, in contrast to the excitatory regions, can be mapped when the base-line responseof these cells is increased by repeated stimulation of the excitatory field(s), while at the sametime a moving stimulusis swept acrossthe entire receptive field. This second stimulus inhibits the responseof the cell when it crossesthe inhibitory regions. Similar results have beenobtained by Watkins et al. (20) who, instead of stimulating in the center of the field, increased
.
. .9
9
I 1.0
. 1.1
1.0
I.1
. 1.2
1.3
4 dbg
CX-type cell.
the baseline firing rate of these normally silent cells by injecting potassiumions near the cell. The easiest way to observe such inhibitory responsesis to take advantage of the relatively high degree of naturally occurring spontaneous activity that one can occasionally find in these cells. One S-type cell of this sort is shownin Fig. 14A. This cell had two excitatory subfieldsfor each direction of movement (S, type), similar to the cell shown in Fig. 7. Stimulus-responsehistograms are shown to two different stimuli. In Fig. MA ,I the responsewas elicited with 0.63’ wide bar moving across the field at 2”/s in one direction. Two excitatory responses can be seen, one to the light and the other to the dark edge. Inhibition is evident before and after each response.In Fig. 14A,2 the bar was reduced to 0.27Oin width, resulting in a summationof the light and dark edge responses.The inhibition to
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
,
SCHILLER,
1300
UNIT 85-8-6.45
I
VOLMAN
+--!I--
0
-
240_
% a
160-
Iz3z -0 80-
0’ ,
,
. I
.2
. 3
.
9
.
.4
5
.6
‘0
d 7
de9
4 6
7 dtg
FIG. 13. Bidirectional CX-type cell with small spatially overlapping light and dark subfields showing a small degree of interaction between direction and contrast.
showsonly excitation to light and dark edges.In the opposite direction, however, a smalldegree of inhibition is evident when the stimulusmoves either side of the excitatory areasis even more over the center of the field. This kind of inhibipronounced. tion was seenin many directional CX-type cells. A detailed study of this cell, using various Another group of cells, which stimuli, yielded the schematicdrawing shown in TONIC CELLS. Fig. 14A,3. It appearsthat each region has its we believe forms a rather distinct third classin own inhibitory side bands. While it was clear monkey striate cortex, comprisesthose neurons that the excitatory light and dark areaswere not which are excited by one sign of contrast and in the same place for the two directions of are inhibited by its opposite. Thesecells, which movement, it was not really possible to deter- were called T-type cells, respondto changesin mine whether this was also true for the inhib- illumination in a relatively sustained(tonic) fashitory side bands. In the S-type cells which we ion and the responseis not contingent on having tested in this manner or with Bishop’s method a contour within the receptive field. Typically (2), such inhibitory regions were frequently evi- they are poorly or not at all oriented, and some fire preferentially to colors without beingsharply dent. Similar examination of CX-type cells failed to selective. They appear to be similar to LGN discloseanalogousinhibitory side bands. Figure cells, but someof them are binocular. Two examplesare shown in Fig. 15. The first 14B shows a direction-selective CX-type cell with moderate spontaneousactivity. Movement of these(15A) is excited by light and inhibited by from left to right, as representedin this figure, dark. The opposite is true for the second cell FIG.
12.
Bidirectional
CX-type cell.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
w x a m
AND
UNIT 90-l-21.60
I
320
3
FINLAY,
MONKEY
UNIT
76-3-21.
STRIATE
CORTEX
I
1301
I5
A
(Fig. 15B); 49 cells in our sample were of this type*
Only a small percentage SELECTIVITY. of the cortical units in our sample were color selective. The characteristics of such cells were similar to those described recently by Dow (5) and earlier by Hubel and Wiesel (lo), and hence were not studiedin detail. Most of them could be classifiedas S-, CX-, or T-type cells, although color-coded S-type cells were especially rare. It was our impression that color-selective cells occur in batches; when we found one in a penetration, succeeding cells were similarly color selective. It is, therefore, possiblethat color is representedin a columnar fashion in striate cortex. COLOR
Shape of receptive field and response to stimuli of various lengths We employed a variety of methodsto assess the size of single-cell receptive fields. When
stationary flashingspotswere used, the majority of cells, including S type, yielded roughly circular or elliptical fields, the long axis of which did not necessarilyfall along the axis of orientation. Receptive-field size can also be plotted with moving stimuli. This works well with movement at right anglesto the axis of orientation. If the stimulus moves at less than about 3”/s, the cell will fire as long as the edgeis within the activating region. Hence the size of the responseis a good index of receptive-field size along this dimension,which we will refer to as the width of the field. Plotting the size of the field along the axis of orientation, which will be referred to as its length, is more difficult with moving stimuli. This may be done by moving an optimally oriented edge(at right anglesto the axis of orientation) progressively toward the center of the field from the periphery. Under theseconditions it is not uncommon that the edge has to be moved beyond the center of the field from either side. Thus one may find that having marked the
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
FIG. 14. Excitatory and inhibitory responses of an S-type and a CX-type cell. A: bidirectional S-type cell with two subfields in each direction. A, 1: response for one direction of movement to a 0.63” wide bar. A,2: response to a 0.27” bar showing a summation of excitatory light and dark regions. A,3: schematic drawing of receptive field. Stipled area shows maintained activity. The response profile of subfields to light and dark edges is shown for both directions of movement. B; unidirectional CX-type cell. B, 1 and 2; responses to single edges in both directions of movement, B,3: schematic drawing of receptive field.
1302
A
UNIT
SCHILLER
., FINLAY,
60-2-11.40 I
UNIT
76-3-21.25
I.
FIG. 15. Two T-type cells. A : excitation to light increment. B: excitation to light decrement.
location of the upper border of the field on the tangent screen as the stimulus was advanced beyond the marked point, creating what mi ht be called a “negative field.” What this probaEly meansis that a certain degree of summationis required along the axis of orientationof the field before a responsecan be elicited (3). Another approach for measuringthe length of the receptive field is to increasethe length of a stimuluscentered in the field in small stepsuntil the responseasymptotes or begins to decline. The size of the stimulusat this point should be equivalent to the length of the activating region. This methodalsoallowsoneto determineto what extent a unit might be inhibited as the length of the stimulus is progressively increased. Such a measurehasoften been usedto place cells in the hypercomplex category (9). In the cat’s and
VOLMAN
monkey’s visual cortex these cells have been shownto have inhibitory regionsalongthe axis of orientation so that once a stimulus had been increasedbeyond a certain lengththe responseis attenuated. Units with such “end stopping” are classedas hypercomplex cells. Simpleand complex cellsin the cat arenot inhibitedin this manner (9), but their summationproperties to stimuli of increasinglengthsappearto be different. Simple cellshave beenshownto summatewhile complex cells are less apt to do so. In order to obtain a measureof receptivefield length, summation, and end stopping, we moved optimally oriented slits of various lengths acrossthe receptive fields of single cells. Most commonly the stimuluslengthsusedwere 0,O. 1, 0.2,0.4,0.8, 1.6,3.2, and 6.4Oof visual angle,the first beinga control no stimuluscondition. These stimuli were swept acrossthe receptive field in a randomized order for 10or 20 trials per stimulus condition utilizing our computer-driven optic display. Stimulus velocity was typically around 2”ls. The results for a strongly stopped unit are shownin Fig. 16in the form of a setof histograms and a curve for one direction of stimulusmovement. This unit had low spontaneousactivity. Responsesrapidly summated with increasing stimuluslength reachinga maximumwith the 0.8 stimulus. With further increases in stimulus length, the responsedeclinedsothat the 6.4”-long stimulus elicited only 15% of the maximum response.This ratio can be usedasan index of end stoppagefor each cell: stopping index = 100 - 6.4” stimulus response x loo best response
Figure 17 shows a representative sampleof nine S-type and nine CX-type cellsstudiedin this fashion from a sampleof 572 units. All the cells shown in the left-hand column of this figure had spatially separate light and dark regions and were, therefore, classifiedas S-type cells. Cells classifiedas CX type had overlapping light and dark regions. We wish to make the following points about this figure: I) The slopesof summation among both S-type and CX-type cells are highly variable and show no consistent differencesfor the two types. Thus, thesecells in the monkey do not seem to differ dramatically in terms of summation. 2) There is considerable variation in the extent to which cells are stopped (index of stoppage at right extreme of each graph). S-type and CX-type cells do not differ in this respect. The relatively continuous variation in the degree of inhibition among these cells suggeststhat the criterion of end stopping may
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
B
AND
MONKEY
CORTEX
I
1303
60-8-12.35
IOSP
-0,?
0 Ids
.I
.
.
.
.2
.4
.8
1
.
I 6
3.2
I
6.4
.I Q STIMULUS
LENGTH
IN
DEGREES
FIG. 16. Response of a strongly stopped CX-type unit with a receptive-field width of 1” to stimuli of increasing length. Histograms are for one direction of movement, reduced from 128 to 32 bins.
not adequately distinguish a separate class of undertaken in an area representinga relatively hypercomplex cellsin the monkey striate cortex. narrow range of the visual field, spanning2”-5” We will examine this question in more detail in from the fovea. Figure 18showsthe distribution the next section. 3) The maximum responseoc- of overall receptive-field widths in this region. curs anywhere from 0.2’ to 6.4”. This would The width of the field was obtained with stimuli suggestthat the length of the receptive field is moving at right anglesto the axis of orientation. often indeed greater than its width, a finding The duration of the moving edge responsewas which does not correspond with that obtained measuredat the half-height point between the using other indexes of receptive-field length. base line and the peak of the response(area containing approximately 75% of the response). Distributions based on For cells having spatially separateresponserequantitative measures gions, field width was measuredto include the The data describedsofar have beenprimarily entire area these regions covered. To assessthe validity of this method we comin the form of individual examples of cell responses. In order to assessthe validity of pared the receptive-field width of 50 units using classificatory schemesand to obtain a general two different measures;the one just described picture of single-cellfunction in visual cortex of andone derived on the basisof flashinga 0.09’ bar the monkey, we examined the distribution of in various parts of the receptive field usingthe computer-driven display. These two measures these cell properties in a large sample. Table 1gives an overall account of the number yielded similar results, showinga correlation of of cells studied and the gross categories into +0.69. This relationship would probably have which they were placed. The criteria used for beeneven better had we beenable to activate the classificationof S-type cellsappear in the legend field reliably with thinner bars, thereby increasof this table. These criteria permitted us to ing the sensitivity of the flash measure. The classify all but 21 S-type cells. This suggeststhat receptive-field sizesderived usingmoving stimuli the cellswe distinguisheddo not form continuous also correspondedclosely with our hand plots. The data in Fig. 18, obtained from 589 cells, distributions in their properties asalready noted. We will next turn to a number of quantitative show a skewed distribution of sizes which vary measuresto determine to what extent the basic over almost a 20-fold range. Only a relatively criteria which had been used in classifying cells smallfraction of this rangeis due to variation in might be applicable for single cells in monkey retinal eccentricity. Even in a vertical penetration, where receptive fields scatter only over a striate cortex. smallarea, field size can vary over a large range, RECEPTIVE-FIELD WIDTH. Our recordingswere with the receptive fields of cells in layer 4 being
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
UNIT
STRIATE
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
MONKEY TABLE
1.
Account
STRIATE
CORTEX
of cells studied
1305
I 127
TOTAL
Total
N 8589
S-type cells CX-type cells T-type cells Unoriented cells Unresponsive cells Unclassified
245 342 49 62 130 297
Total
1,125 S cell
breakdown
1. Response to one edge a. Unidirectional, L only b. Unidirectional, D only c. Bidirectional, L only d. Bidirectional, D only 2. Response to two edges a. Unidirectional b. Bidirectional c. L in one direction, D in d. L in both directions, D e. D in both directions, L 3. Multiple responses a. Three fields b. Two opposite direction 4. Uncategorized
28 28 6 5
.4
.5
.6
.7
xl
.9
IO
II
1.3
14
I.5
1.6
I7
I.8
I9
2.0
I.6
I.7
I.6
I.9
2 0
1.6
I7
I.6
I.9
2.0
I.7
I.6
I.9
20
ce IIS
type N=l60
47 21 36 14 17
other in one in one
12
17 5 21
fields
Total
245
Criteria for S-type cell classification (calculations for these criteria are described in RESULTS, Distribution bused on quantitative measures. 1: contrast measure is less than 15 (response to dark edge only) or more than 85 (response to light edge only) (S, type). 2a: two separate subfields where 15 < C < 85, interaction index is greater than 10, overall directionality is less than 50, and the difference between directionality for the light edge ( Di) and that for the dark edge ( DD) is less than 50 (S, type). 2b: Two separate subfields, 15 < C < 85, I > 10, overall directionality more than 50, and 14. - D,,I < 50. (S, type). 2c: I s 10, 15 < C < 85, DL and DD are both G 30 in opposite directions (S, type). 2d, e: 15 < C < 85, I > 10 and 1 D1, - DI,I s 50 or DL(DD) > 30 if Q,( D,) < 10 (S4 type). 3a: Three subfields alternating light and dark in at least one direction (S, type). 3b: Two separate subfields where each meets the criteria for 2c (& type). Unclassified: not enough data to classify but enough separation of subfields to be called S-type cells.
cx
type
cells
N=262
.8
IO
-
.9
IO
I I
I2
I3
I4
Unotiented
9
IS
cells
N=39
.I
.2
.3
.4
.S
.6
.7
.B
.9
I.0
RECEPTIVE
FIG.
18.
Distribution
I.1 FIELD
of
I2
I3
I4
IS
1.6
SIZE
overall
receptive-field
width.
S-type and CX-type cells can be differentiated. This can be expressed in terms of a potential misclassification between these classes. Choosing the best criterion to differentiate S-type from CX-type cells on the basis of subfield size results in a misclassification of 18% of the population. Using the same criterion for overall receptivefield size, 34% of the cells are misclassified. SPATIAL
SEPARATION
BETWEEN
SUBFIELDS.
CX-type cells and many S-type cells respond to both light and dark edges. By defini-
tion, the prime distinction between these classes of cells is that the subfields of S-type cells are spatially separate, while those of CX-type cells are spatially superimposed. To check on the validity and the classificatory power of this distinction we compared the extent of spatial overlap between the light and dark subfields in S-type cells with bipartite fields and in CX-type cells. This was done by using the measure of receptive-field size and the center to center spatial separation between the subfields. The follow-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
1. 2. 3. 4. 5. 6.
distribution
1306
SCHILLER,
FINLAY,
67
s type N=
type
cells
N=258
FIG. 19. Distribution of subfield widths for S-type and CX-type cells. The light-edge receptive field was used unless the dark-edge response was better.
ing formula was used to determine the extent of separation or overlap between the subfields: overlap = 1/2L + 1/2D + center-to-center l/2L + 1/2D - center-to-center
field separation field separation
where L standsfor the receptive-field size of the light-edge responseregion and D standsfor the receptive-field size of the dark-edge region. A
value of 100obtained this way is equivalent to a 100%overlap betweenthe light anddark fields. A value of 0 indicates that the two subregionsare immediately adjacent without overlapping. A negative value indicates a space between the subfieldsas determined by this measure. Figure 20A shows the comparison between S-type cells with bipartite fields and CX-type cells. Thesedata showthat our criterion successfully discriminatesthese two types of cells, with only a few falling into a common area. A few S-type cells had subfieldswith an unresponsive region betweenthem. This is somewhatexaggerated in Fig. 20 becausereceptive-field size was measuredso as to include only 75% of the responsearea. For example, the value for the cell in Fig. 4 was -18 although it is clear that there is indeed a small degree of overlap between the subfields. By changing the receptive-field size criterion to the less reliable but more encompassinglevel of 90% of the responsearea, one would effectively shift the 0 point for S-type cells to the -20 region on the figure. This still leaves a number of cells which show an unresponsive area between the subfields. These data suggestthat the prime criterion usedin differentiating S-type and CX-type cellsis an effective one. Our analysisshowsthat choosing the best criterion to differentiate S-type from CX-type cells on the basisof subfieldseparation results in misclassification of only 4% of the B
A
60 -
s type cells N=ll2
?
SUBFIELD
SEPARATION
z 3 IA 0
~
20-
: r z 40.
CX type cells N=l84 60 -
FIG. 20. A: spatial overlap between light- and dark-edge response areas for bipartite S-type and CX-type cells. B: center-to-center separation of subfields of bipartite S-type cells.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
CX
cells 144
AND VOLMAN
MONKEY
STRIATE ICO
1307
I
143
TOTAL N=762 00 -
20
m 60t z 3
30
40
!iO
60
70
8-O
90
100
50
S type cells
IO
20
30
40
50
60
70
80
90
100
CX type
cells
N= 313
DIRECTIONALITY. To determine the distribution of direction specificity, for each cell a count was made of the total number of responseselicited by moving edgesin each direction to both signsof contrast basedon 60 trials. An index of direction specificity wascalculatedby the following formula:
index of directionality
= ;z;
Ex
loo
where the smallernumber was always placed in the numerator. This measuredoes not reflect anything about absolute direction in the visual field but, instead, specifiesthe directionality of a cell to opposite directions of movement at right anglesto the axis of orientation. A small value reflects a tendency for the cell to respondonly in one direction of movement; such cells are thus consideredpredominantly unidirectional. A high value, such as 90-100, indicates that the cell respondedequally well to both directions and, hence, could be called a bidirectional cell. Figure 21 shows the distribution of direction specificity for 762 cells, of which 220 were classified as S type and 3 13 as CX type. The CX-type cells show a roughly bimodal distribution, suggestingthat unidirectional and bidirectional cells may be considered as two separate subpopulationsof CX-type cells. The distribution for S-type cells is skewed, strongly favoring unidirectionality. Sincethis sampleincludescells which show an interaction between responseto direction and to contrast (S, type), we attempted
IO
U
20
30 INDEX
40
50 OF
60
70
80
90
100
B
DIRECTONALITY
FIG. 21. Distribution of directionality. The index of directionality is 100 times the total response in the worse direction divided by the total response in the better direction. An index of 100 indicates a bidirectional cell; an index of 0 indicates a unidirectional cell.
to analyze the extent of directionality by reducing this confounding variable. We did this in two ways. First, we analyzed all single-contrast,S,type cells separately. Figure 22A shows this analysisfrom which it is clear that thesecellsare predominantly unidirectional. Second, we determined for all S-type cells the extent to which either a light or dark edgealoneelicits a unidirectional response.The distribution derived in this manneris given in Fig. 22B. It showsthat when the interaction between contrast and direction is reduced asa variable, the unidirectional properties of S-type cells are more apparent. This kind of analysisdoesnot affect the distibution of complex cells as given in the previous figure because thesecells (seealso Fig. 24) showno interaction between contrast and direction. .AST. To assesshe overall selectivity of striate cellsfor contrast we counted the number CONTR
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
population. The overall differentiation of these two classesof cells may actually be even better, since this sample does not include the easily categorized single-contrast, S-type cells. In Fig. 20B we plotted the center-to-center subfieldseparationof S-type cellsin degrees.The majority of cells had 0.2-0.3Oof center-to-center separation. The average spatial separation is about as great as the average receptive-field size of each subregion(seeFig. 19).CX-type cellsare not shownin Fig. 20because,with the exception of three cells, they would all falI into the O-O.1 column. We were also interested in determining for S-type cells the relationship between the size of the subfieldsand the degree of center-to-center separationbetweenthem (measuredin degreesof visual angle). A correlation performed between thesetwo measuresusing 135S-type cellsdid not show a significant relationship(Y = 0.15). Thus it appearsthat in our sample,which was restricted to a region 2”-5” from the fovea, the size of the subfieldsin bipartite field S-type cells and their spatial separationsvary independently of each other.
CORTEX
1308
SCHILLER,
FINLAY,
AND VOLMAN
TOTAL 160
-
N= 654
20
30
40
50
60
70
80
90
100 10
20
30
40
50
60
70
80
90
100
S type
cells
N=213
IO
20
30
40
50
60
70
80
90
100
W z 3 L
124 I20
-
100
-
80
-
60
-
CX type
IO
20 INDEX
30
40 OF
50
60
70
80
90
100
DIRECTIONALITY
Index of directionality for S-type cell subfields. A: units which respond to only one sign of contrast. B: directionality index distribution of single contrast subfields considered separately. The proportion of unidirectional fields increases when light and dark response subfields are considered separately. FIG.
cells
N=308
22.
of responses elicited by the light and dark edges separately, independent of direction of motion. These data were obtained, as described earlier (Figs. l-13), by moving rectangular stimuli or single edges across the receptive field at right angles to the axis of orientation. An index of the contrast specificity of a cell was derived by the following formula: contrast specificity = XL+CL CD ’ loo where L and D stand for the responses elicited by the light and dark edges. An index value of 10 or less means that the cell responded almost exclusively to a light edge, while a value of 90 or more indicates a strong preference for dark edges. A score of 50 indicates an equal response strength to light and dark edges. In Fig. 23 the data based on 654 cells show
.O
20
30
40
50
60
70
80
90
IO0
Unoriented
cells
N =50 ‘0
20
30
40
50
60
OF
CONTRAST
I
70
80
90
100 D
INDEX
FIG. 23. Distribution of contrast dependency for total, S-type, CX-type, and unoriented units. The index of contrast = (response to light edge/response to both edges) x 100. An index of 100 indicates response to light only; a unit responding equally to light and dark has an index of 50; a dark response only gives an index of 0.
that the majority of cells in our sample responded fairly equally to light and dark edges. CX-type cells as well as unoriented cells show this tendency most strongly. By contrast, S-type cells are evenly distributed on the scale, with
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
lo 131
MONKEY
STRIATE
CORTEX
1309
I
many cells respondingonly to one sign of contrast. Not included in this sampleare the tonic cells, all of which respondin an excitatory fashion to only one sign of contrast. INTERACTION TION. We
BETWEEN
CONTRAST
AND
DIREC-
100
m Ir 3 L 0
a w
-
s type N
cells
8211
50 -
IO
20
30
40
50
interaction = JL+ + D+1 Or [L+ + D+l x 100 total L + D The smaller combination was used in the numerator. This measure ranges from 0 to 50, where a value near 0 reflects very stronginteraction of the sort shown for unit 91-I-22.27 in Fig. 5, while a value near 50 would indicate no interaction. Figure 24 showsthe distribution of interaction valuesfor 689cells, of which 211were classified as S-type, 301as CX-type, and 50 as unoriented cells. These data show that CX-type cells, as stipulated, do not have a significant degree of interaction. S-type cells show some bimodality in their distribution, which is indicative of two subgroups, one showing a low and the other a high degree of interaction. lNHIBITION LENGTH.
WITH
INCREASING
STIMULUS
The measurefor inhibition along the axis of orientation was obtained by sweeping bars of different lengths across the receptive field, as already described. Stopping index (stimulus length-responseratio) was calculated by dividing the response obtained with the longest stimulus(6.4”) by the optilmalresponse. This fraction was multiplied by 100 and this value was subtracted from 100. If the cell’s maximum response was elicited by the 6.4” stimulus, a value of 0 was assignedto that unit. Thus, a low value on this scale indicates an unstopped unit, that is, one which responds without being inhibited by increasingthe length of the stimulus. A value near 100indicatesa cell which is strongly inhibited by stimulation of regionsalongthe axis of orientation away from the receptive-field center. The distribution for end stoppageshown in Fig. 25 was determinedfrom the analysisof 572 cells, of which 129were S-type, 226-CX-type, 78 INDEX
OF
INTERACTION
FIG. 24. Distribution of interacti on between contrast dependence and directionality. The index of in-
teraction (1) = (L-+ + D+- or L+- + D--+/total L D) x 100. An I of 50 shows no interaction; a unit with an I of 0 responds to light and dark edges in opposite directions.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
have shown that somes-type cells exhibit a strong interaction betweencontrast and direction. To measurethe distribution of this interaction, the samedata, obtained by moving light and dark edges over the receptive field, were reanalyzed to obtain an index of interaction. This was computed for each unit usingthe following formula:
1310
SCHILLER,
60- TOTAL N=572
60-
80
we S type
70
60
50 40
30 20
IO
cells
N = 129
10090 v)
80 70
60
50 40
30 20
IO
60-
+ 5 3
50- CX type cells N=226 40-
100 90 80
70
60
50 40
30 20
IO 21
20-
Unoriented
cells
N=70 IO-
100 SO
80
70
60
50
40
30
20
50 40
were unoriented, and 19were T-type cells. The distribution appears to be primarily unimodal with no clear difference between the classes.It is our impressionon the basisof thesedata that this measure cannot be used to separatecells reliably into a hypercomplex category in the monkey. In order to obtain an idea about the extent to which an analogueto end stoppingis observable in the LGN, we also studied 16cells in the same 2”-5’ area of the visual field using similar methods. The index of stoppagefor thesecells was15, 21,22,23,27,33,38,39,43,45,49,63,65,74,75, and 100.It may be saidthat mostLGN cellsusing this method show pronounced surround inhibition and, hence, demonstrate strong end stopping. In what senseend stopping representsa cortical transformationthen isa problem. We will consider this question in the DISCUSSION. SPONTANEOUS ACTIVITY. The last distributional measuredealswith spontaneousactivity, which in previous work has been shown to be lower for simple cells than for complex onesin the cat (1). The meannumberof dischargesoccurring with no stimulus in the receptive field was measured during the testing of various stimuluslengths. A l-s time period was sampled randomly during these trials. The distribution for spontaneousactivity is given in Fig. 26. It showsthat S-type cells have lower spontaneousactivity than CX-type cells (meanfor S = 1.2, for CX = 4.9). Yet there is a considerableoverlap in population. Particularly evident is the fact that there are many CX-type cells which have low spontaneousfiring rates. Choosingthe bestcriterion to differentiate S-type from CX-type cells on the basisof spontaneous activity results in misclassificationof 27% of the population.
Distribution of recep tive--eld properties of cells in cortical layers Using lesions to mark electrode tracks, we identified the location of 201 cells in striate cortex. Figure 27depictsthe reconstruction of part of a tangential penetration between two marker lesionsin which 18 isolated single cells were recorded. Two cells were not held long enoughto
IO
lo- T type cells
10090 80 70 60 STOPPED STIMULUS LENGH
AND VOLMAN
30 20 IO UNSTOPPED RESPONSE RATIO
FIG. 25. Degree of inhibition as a function of increasing stimulus length. The measure of end stoppage (stimulus-length response ratio) was obtained by sweeping the receptive field with bars or edges of various lengths and applying the following formula: 100 - (response to bestlength/response to longest bar) x 100. Largevaluesindicatestronglystoppedcells.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
10090
FINLAY,
MONKEY
STRIATE
CORTEX
I
1311
edge responseis represented in the schematic drawings. The short, heavy line in the center of each schematic drawing representsthe axis of orientation. The stimuli used to generate the data from which the drawingswere derived were 300 moved acrossthe receptive field at right angles to the axis of orientation, as described earlier. The markingsare in 0.1’ steps. The main points to be consideredabout this figure are the following: I) Both S-type and CX-type cells can be 200 found in layer 6. This is also true for T-type cells. 2) Receptive-field size varies over a broad range amongthe cells in the two layers sampled in this pass.3) Nearby cells have similar orientations (12) but their directions can be 180” reversed. This is especially evident for the cells numbered 15 and 16. These two cells were recorded from simultaneously.4) The double-field &-type cell (7) hasa small center-to-center separation, which with stationary, flashing stimuli + 2 8 16 32 might not have beendiscernible. This cell is only marginally different from a small CX-type cell. 116 This cell falls into the overlapping region of Fig. 20. 5) In this penetration there are four singleS type cells contrast, S-type cells in layer 6 (numbers6, 9, N= 137 15, and 16). Figure 28 showsthe overall distribution of cell types in the cortical layers (11). S-type cells were found in all layers, and in layer 4c most of the isolated cells recorded were S type. This layer containspractically no CX-type cells, which are numerousin all other layers. The samplein Fig. 28is basedon only 201cells and does not include data from a number of penetrations, not reconstructedhistologically, in I6 32 + which we estimated the location of the cells in cortex by noting when the electrode touched brain and when it entered white matter. The CX type cells relative thickness of layers in striate cortex is quite constant, with layers 1,2, and 3 constituting 81 N=245 the top 35% of tissue, layers’ 4a, b, and c, the next 30%, and layers 5 and 6, the remaining25% of gray matter. We could, therefore, assigncells with a reasonabledegree of accuracy to these three subdivisions. The locations of 295 cells were determinedin this way giving a total of 496 cells whose approximate positions in cortex were known. The number of S-type cells in layers l-3, 4, and 5-6 were 29, 36, and 35, re16 32 + 2 8 spectively; there were 62, 48, and 50 CX-type SPIKES PER SECOND cells in these subdivisions of cortex. Of the Distribution of spontaneous activity for single-contrast, S-type cells there were 8 in FIG. 26. total, S-type, and CX-type cells. layers 1-3, 8 in layer 4, and 7 in layers 5-6. T-type cells had the following distribution: be identified. Two were classifiedas tonic types layers l-3, 3; layer 4, 7; layers 5-6, 7. In Fig. 29 distributions among the cortical (T), five as S type (6, 7, 9, 1.5,16) and eight as CX type. The receptive-field organization of layers are shown for receptive-field size, inhibithese cells in terms of optimal orientation, direc- tion along the axis orientation, and spontaneous tionality , receptive-field size, and light-dark activity. 400
-
381
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
u
1312
SCHILLER,
FINLAY,
AND VOLMAN
2 3 40 4b
14
13 I2
II
10987654
3
2
I
11
FIG. 27. Tangential penetration reconstruction (layers S and 6). Filled circles show position of marker lesions. Receptive fields were mapped with moving edges or squares and are shown schematically marked in O.l” steps; the short, heavy lines are the axes of orientation. Cells number 6, 7, 9, 15, and I6 are S type; 5 and I I, T type; all others are CX type. The two units between 10 and II were not held long enough to classify.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
16 I5
MONKEY Number
20
ii )r
CORTEX
1313
I Percent
cells
60
-
40
-
60
-
40
-
20
-
20
-
of
cells
-
-
IO -
0 W
tz W z
IO
-
2-3
40
4b
4c
6 CORTICAL
2-3
40
4b
4c
LAYERS
FIG. 28. Distribution of cell types in cortical layers. Locations of 201 cells were determined from histological reconstructions. Graphs at left show numbers of each type cell found in the various layers. Graphs at right show the percentage of cells in each layer classified into the various cell types.
These results indicate that: I) The receptivefield size of S-type cells is the same in all the layers, except for layer 4c (not shown separately in Fig. 29for which 4a, b, and c were combined), where all the cells had fields smaller than 0.2”. Receptive fields of CX-type cells are smallest in layers 1-3, and get progressively larger in deeper layers. 2) The degree to which cells are inhibited along the axis of orientation varies considerably with depth. Cells in layers 1-3 tend to be more strongly stopped than cells in layers 5 and 6, where most cells are unstopped, while layer 4 falls between these extremes. The differences between S-type and CX-type cells are small. 3) S-type cells have low spontaneous activity in all
layers of cortex. CX-type cells also show low spontaneous activity in layers 1-3, but tend to have higher rates of maintained firing in layers 5-6. This measure seems reasonably effective in discriminating between S-type and CX-type cells in layers 5 and 6, but is ineffective in the other layers. In contrast to the overall misclassification of 27%, in layers 5-6 only 11% of the cells are misclassified using spontaneous activity as a criterion.
Alert monkeys We were concerned in the course of our recordings that the state of the animal, including anesthesia, might produce sufficiently pro-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
20
of
STRIATE
1314
SCHILLER, FIELD
FINLAY,
AND VOLMAN STOP
SIZE
N= 117
SPONTANEOUS
N=62
75-
75-
50-
50-
25-
N= 103
N=225
25-
cn
.2
.4
.6
.8
1.0
t
FlEL,D SIZE
.2
.4
IN DEGREES
.6
.8
1.0
+
100
LAYERS
80
60
40
INDEX
20
loo
80
60
40
20
OF STOPPAGE
I
LAYERS
4
8
I6
SPIKES
t
I
PER
4
8
I6
+
SECOND
FIG. 29. Field size, inhibition along axis of orientation, and spontaneous activity as a function of depth for S-type and CX-type cells. Depths were obtained from reconstructions and the estimation method described in the text. CX-type cells become larger, less stopped, and more spontaneously active with increasing depth; S-type cells show a change only in their stopping index.
nounced changesin striate cortex to render the properties of singlecells in our situation different from those of the alert, normally behaving monkey. Our paralyzed animals were anesthetizedinitially by Pentothaland subsequentlyby N20. We did not seeany striking differencesin the properties of cells using these two agents. The spontaneousactivity and responsivenessof cells was somewhatlesswith Pentothal, but this was most evident only for the first 5-10 min following readministration. Irrespective of the anesthetic agent used, the responsivenessof cells in the superficial layers (but not in layers 4, 5, and 6) tended to decrease with time, especially during the latter part of the second day of recording. Becauseof theseobservationswe undertook to record from two alert monkeys which had one eye surgically immobilized as described in the METHODS section. Only qualitative data were obtained in these animals. Our recordingsfrom 109 cells suggestthat the properties of units in these animalswere quite similar to those of the paralyzed, anesthetized animals. In the superficial layers many cells respondederratically and tendedto habituate, while the cells in layers 4, 5, and 6 dischargedconsistently to visual stimuli. The majority of cells were oriented although this sample seemedto contain a somewhat greater percentage of unoriented cells. We also found fewer S-type cells, which may have been due to the fact that the recording stability and time spent
in a given region was lessthan in the paralyzed animals. In general, our qualitative observations suggestthat the properties of singlecellsin anesthetized, paralyzed animalsare representativeof the alert monkey’s visual cortex. DISCUSSION
This section discussesthe various types of cells described in the RESULTS and the significance of the distributional measures.Possible models for cortical cells will be considered briefly.
Characteristics striate cortex
of single cells in
The resultssuggestthat a variety of S-type cells may be distinguishedin the striate cortex of the monkey; their properties are summarized schematically in Fig. 30. The singlecontrast, S-type cell in this figure shows the simplest type of oriented cell, which exhibits a singleexcitatory region to moving edges,within which it respondsonly to one sign of contrast change. In some of these cells it is possibleto demonstratean antagonistic surroundby placing a stationary stimulusin the center while flashinga large, superimposedstimulus. These cells are striking in that practically all of themare unidirectional. The secondnotable group is the double-field, !$ cell, in which one subfieldis excited by a light edge and the other by a dark edge. Both fields S-TYPECELLS.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
l-3
l-3
MONKEY
2
3
OF
DEGREES
1315
I
4
VISUAL
OF
CORTEX
ANGLE
VISUAL
AhSLE :6
.I
2
.3
4
5
6
7
6
.9
dsQ
S .I
.2
.3
.4
:6
:5
7
6
9
1.0
de9
6 -
O'JQ
I
FIG.
2
3
4
5
6
7
8
9
IO
I I
I
2 , deQ
30. Schematic drawings for seven S-type cells and one CX-type cell.
respond to movement in the same direction, and two-thirds of such cells in our sample were completely unidirectional. The third group are cells which show an interaction between contrast and direction of movement (S,). They have two subfields, and
each responds only ment, but in opposite The fourth group S-type cells are those elicits responses for ment, while the other
to one direction of movedirections to each other. of commonly observed in which one contrast edge both directions of movecontrast edge elicits a re-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
I DEGREES
STRIATE
1316
SCHILLER,
FINLAY,
VOLMAN
mean receptive field of the single-contrast, S, cells in our sample is 0.200 (SD = 0.09); that of the double-field, S, cells is 0.40 (SD = 0.10). The proposed combinations do not go contrary to the known cytoarchitecture of cortex: while orientation is strictly maintained in a columnar fashion, no such organization is evident for direction; within a column reversals of preferred direction are common. Furthermore, one can, on occasion record from two single-contrast, S, -type cells simultaneously, as already noted, which have spatially separate fields and are sensitive to opposite directions of movement. The problem with any sort of hierarchic concept is that recordings from S-type cells, including S-type, are obtained in all cortical layers. One must, however, consider the possibility that the order of cell types in these layers is greater than revealed by extracellular recording methods. If single-unit records were obtained not only from cell bodies, but to some extent also from axons and dendrites, such order would be significantly obscured. In this respect it is interesting that progressive changes in receptive-field properties as a function of recording depth in cortex are found only in CX-type cells. If unit signals are indeed recordable at sites other than near the cell body, the probability of this would be higher for S-type cells provided they have the profuse axonal terminations within striate cortex postulated by the Hubel-Wiesel (8) hierarchic model. The proposed hierarchy is only one possible way of conceptualizing the various S-type cells since models of cells can also be constructed assuming no such hierarchy. A more detailed discussion of models of visual cortex will be deferred until the fifth paper in this series. Irrespective of the manner in which these cells are constructed, the question remains as to how a single cell can have spatially separate excitatory subfields of the sort described. Since input from the LGN seems highly ordered topographically (12), one possibility is that inputs converge from spatially separate cortical areas, from one or two hypercolumns away where cells of the same orientation but of a slightly different spatial location reside. The other possibility is that the different layers of the LGN, consisting of repeated representations of the visual field project into visual cortex with some misalignment. These projections form multiple layers of terminal fields (6) and thus could be the source of the spatially separate subfields observed in S-type cells. If this is the case one might expect that in some cases the subfields have their origin in different layers of the LGN. This notion does not explain one interesting problem, namely that there are potential combi-
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
sponse only for one direction of movement. These cells were called U-B or S4 cells. More complicated arrangements are observed in S-type cells which have several subfields (S,, S,, and S,). Some of these cells-may have three distinct areas with, typically, a center region responding to one sign of contrast change flanked by two regions responsive to the opposite contrast (S,). In some cases these cells are completely bidirectional, while others may demonstrate unidirectionality for some regions and bidirectionality for others. The properties of the various S-type cells we discussed are similar to the simple cells described by Bishop, Coombs, and Henry (1) in area 17 of the cat. In Fig. 10 of their paper these investigators show a variety of spatial arrangements of simple-cell subfields. We have seen cells similar to almost all of these. It is rather interesting that this similarity in cat and monkey cells extends, in part, to the percentage of each cell type; in the sample of Bishop et al., 9 of 43 cells responded only to a single edge, and all of these cells were unidirectional. Of the cells responding to 2 edges, 22 were unidirectional and 7 bidirectional, again a ratio rather similar to the one we obtained in the monkey. However, we have seen more cells showing interaction between direction and contrast (S,) of which Bishop, Coombs, and Henry report two. The U-B cells (S,) were also more common in our sample. How might the more elaborate receptive-field arrangements of S-type cells be derived from their inputs? While a variety of possibilities exist, the array of S-type cells described might well be understood by assuming a hierarchic relationship among them. Lowest in the hierarchy are the single-contrast, S, cells, which may be thought of as being first-order S-type cells. Assuming such a hierarchy, the initial transformation of LGN input in visual cortex involves not only orientation, but also direction selectivity: it appears that at this early stage a conversion occurs which renders almost all cells unidirectional. S-type cells with more complicated properties may be constructed by convergent excitatory input from the first-order S, cells. Thus a double-field, & cell may result from two singlecontrast, S, cells of opposite contrast response, but with the same orientation and direction. An S3 cell may be constructed in the same manner, but the two single-contrast, S, cells would have direction preferences 180’ in opposition. Cells with multiple subfields may be made from a variety of combinations. In favor of such a hierarchic relationship among S-type cells is the fact that the receptive-field width of these cells increases with increasing complexity. This is especially clear when one compares the S, - and $,-type cells. The
AND
MONKEY
STRIATE
CELLS. We defined CX-type cells as thoseoriented cells which have a unified activating region within which a responsecan be elicited to both light increment and light decrement. The receptive fields of these cells tend to be larger than those of S-type cells, although there is some overlap in the distribution obtained. The measure of direction selectivity suggeststhat these cells constitute two subgroups: unidirectional and bidirectional cells. CX-TYPE
CELLS. The T-type cells, which appear to form a distinct group, are characterized by the fact that they respondto visualstimuli in a relatively sustainedfashion and are excited by only one sign of contrast. They are generally poorly oriented, or not oriented at all, and a higher percentageof them are color specific than S-type and CX-type cells. Thesecells may be analogous to the class I cells reported by Dow (5). Since their properties are somewhat similar to LGN cells, they do not appearto transform the geniculate input greatly. Therefore, it is likely that these cells are driven primarily by geniculocortical afferents. T-TYPE
Quantitative
measures
The basic criteria which have been used in earlier studiesto distinguishamongcells types in cat visual cortex are the following: I) Simple cells have one or more spatially distinct activating regions which can be plotted with stationary flashes. Complex cells do not show such separation. 2) The responseswithin eacharea summate with increasingstimuluslength in simplecellsbut not in complex. 3) Low spontaneousactivity is a property of simplecells. 4) Simple cellsare most numerous in layer 4, while complex cells are found primarily above and below this layer. 5) Simplecells have relatively smallreceptive fields in comparison with complex cells. 6) Hypercomplex cells form a separateclassdistinguished from others by virtue of the fact that they are strongly stopped for stimulus length along the axis of orientation. The distribution of cell properties and the histological data make evident the following points: I) The measure of subfield overlap (Fig. 20) shows a bimodal distribution suggesting two
I
1317
classesof cells. Only 36 cells appear in an area commonto both classes.2) Summationfor stimulus length, as shown in Fig. 17, doesnot differentiate S-type from CX-type cells. 3) The distribution of cells according to spontaneousactivity is unimodal. Cells classifiedas Stype and CX type show a considerable overlap so that this measurediscriminatesthese two classesonly in layers 5 and6.4) There isextensive intermingling of S-type and CX-type cells in almostall layers of cortex. 5) Receptive-field width shows a broad distribution, which appearsto be unimodal. CXtype cells have, on the whole, larger overall receptive fields than S-type cells, but there is an overlap of field size for a considerableportion of the two populations. A better differentiation between these classesis obtained, however, when subfield size to one sign of contrast change is compared. These observations suggestthat with the measuresdetailed, S-type and CX-type cellscan best be distinguishedon the basisof the spatialseparation of the light and dark responseareas. This measure does not, however, establish unique classes. Our qualitative observations confirm that Stype cells have inhibitory sidebandswhile CXtype cells lack this property. S-type cells with more than one excitatory subfield seemto have inhibitory sidebandsfor each of their subfields. Since mapping of these regions is difficult, it cannot be usedasan efficient meansof classifying cells into the S and CX categories. We have not studiedthe inhibitory property in sufficient detail to make quantitative and distributional statements about it. The major problem we encountered in our classification was with the third memberof the hierarchy as originally proposed by Hubel and Wiesel (9): the hypercomplex cell. While it is clear that examples of cells with hypercomplex properties can be found in area 17in the monkey, the strength of end stopping does not seemto provide a satisfactory criterion for classifying cellsin the monkey. The extent to which one can observe inhibition along the axis of orientation forms a unimodal, skewed distribution. Many, although not all, of the moderately and even strongly stoppedcellscould be clearly classifiedas S or CX type. End stopping, however, is not evenly distributed in the cortical layers. Significantly more cellsare stoppedin layers l-3 than in layers 5-6. This is an agreementwith the reports by Hubel and Wiesel (10) and Poggio(15). On the basisof this one might suggestthat the hypercomplex attribute is the product of a general inhibitory network, the effectiveness of which diminishesfor progressively deeper cells in cortical gray matter.
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
nations of subfieldsin S-type cells which do not appear. For instance, cells in our samplewhich have two or more spatially separate fields responding to the same sign of contrast are extremely rare or nonexistent, although it is true that such combinationsmight be difficult to recognize. In spite of this, it is fair to say that some combinations have a high probability of occurrence while others have a low probability.
CORTEX
1318
SCHILLER,
FINLAY,
VOLMAN
a specific analysis of the visual input and 2) that they selectively project to other brain sites, one could perhaps better understand their function. In the subsequent papers we will attempt to contribute to the first question. Assessment of the second awaits further work. The major findings in this study and the inferences drawn from them may be summarized as follows: a) Several different types of S-type cells may be discerned in monkey striate cortex. The subfields of these cells appear to be produced by excitatory action. b) A hierarchy may exist among S-type cells. c) S-type cells are predominantly unidirectional, which suggests that the cortical transformation for directionality may occur at an early stage in striate cortex. d) S-type units, including the single contrast, S, cells are found in all cortical layers. c’) The only measures used in this study which yield well-separated distributions for S-type and CX-type cells are those which specify the spatial organization of the light- and dark-response areas of the receptive field. f) Unidirectional and bidirectional CX-type cells represent two distinct classes. g) Tonic cells, which are excited by only one sign of contrast change, form a third class of cells. h) The measure of the inhibition along the axis of orientation forms a continuous distribution which, therefore, does not yield a separate class of hypercomplex cells in the monkey’s striate cortex. ACKNOWLEDGMENTS
The authors thank Andres Polit, Cynthia Richmond, Kathy Anderson, and Louis Porter for their help on this project. This research was supported in part by grants from National Institutes of Health (EY00676) and the Alfred P. Sloan Foundation (72-4-l).
REFERENCES 1.
2.
3. 4.
5.
BISHOP, P. 0, COOMBS, J. S., AND HENRY, G. H. Responses to visual contours: spatiotemporal aspects of excitation in the receptive fields of simple striate neurones. J. ph~~iol., London 219: 625-657, I97 I. BISHOP, P. O., CUOMBS, J. S., AND HENRY, G. H. Interaction effects of visual contours on the discharge frequency of simple striate neurones. J. Physiol., London 2 19: 659-687, 197 1. BISHOP, P. 0. AND HENRY, G. H. Striate neurons: receptive field concepts. Invest. Ophthulmol. 11: 346-354, 1972. BROOKS, B. AND JUNG, R. Neurophysiology of the visual cortex. In: Handbook of Sensory Ne rrrop h ys iology , edited by R. Jung. Berlin: Springer, 1973, vol. VII/3B, p. 325-440. Dow, B. M. Functional classes of cells and their laminar distribution in monkey visual cortex. J.
Ncurophysiol.
37: 927-946,
1974.
6.
HARIING, J. K., BADE, B. J., AND ATENCIO, E. W. The organization of retino-geniculate and geniculostriate pathways in the grey squirrel (Sci-
urus curolinesis).
7.
Am.
Assoc.
Anat.,
87th,
1974.
D. H. AND WIESEL, T. N. Receptive fields of single neurones in the cat’s striate cortex. J. Physiol., London 148: 574-591, 1959. 8. HUBEL, D. H. AND WIESEL, T. N. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex. J. Physiol., London 160: 106-154, 1962. 9. HUBEL, D. H. AND WIESEL, T. N. Receptive fields and functional architecture in two nonstriate visual areas ( 18 and 19) of the cat. J. Neurophysiol 28: 229-289, 1965. 10. HUBEL, D. H. AND WIESEL, T. N. Receptive fields and functional architecture of monkey striate cortex. J. Physiol., London 195: 215-243, 1968. HUBEL,
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
The extent to which end stopping represents a transformation of the L,GN input to visual cortex is a knotty problem. It appears that the majority of cells in the LGN are strongly inhibited by their surrounds when long bars or edges are used, more so than the population of cortical cells we studied. On the basis of this one might consider the possibility that the significant cortical transformation produces a lack of end stopping rather than its opposite. This may be produced, as suggested by Hubel and Wiesel (8), as a result of a great deal of convergence of LGN cells on cortical cells. Strongly stopped cells in cortex may be produced either by a sparse input from the LGN or by a special inhibitory circuit in cortex. A more thorough study is needed to assess the nature of this transformation. Finally, we might pose the question of what the functional significance of the various cell types might be. Why, for example, are so many S-type cells comprised of bipartite or multipartite excitatory fields? Does this arrangement permit a specific form of analysis of visual information? Assuming a hierarchy, with S-type cells providing the input to CX-type cells, why should this property be promptly lost in CX-type cells? Perhaps S-type cells are involved in the analysis of spatial frequency or some other parameter and send this information to another structure for further analysis. On the other hand, it is also possible that the spatially separate subregions observed in these cells serve no specific function, and may be an outcome of a relatively sparse input from the slightly misaligned regions of the multilayered I, GN terminal fields. Further pooling of S-type cell outputs would eventually get rid of this defect. If it could be shown that I) S-type cells perform
AND
MONKEY 11.
12.
13.
15. 16.
HUBEL., D. H. AND WIESEL, T. N. Laminar and columnar distribution of geniculocortical fibers in the macaque monkey. J. Comp. Neural. 146: 421-450, 1972. HUBEL, D. H. AND WIESEL, T. N. Sequence regularity and geometry of orientation columns in monkey striate cortex. J. Comp. Neurol. 158: 267-294, 1974. HUBEL, D. H. AND WIESEL, T. N. Uniformity of monkey striate cortex: a parallel relationship between field size, scatter and magnification factor. J. Comp. Neurol. 158: 295-306, 1974. MAFFEI, L. AND FIORENTINI, A. The visual cortex as a spatial frequency analyser. Vision Res. 13: 1255-1267, 1973. POGGIO, G. F. Spatial properties of neurons in striate cortex of unanesthetized macaque monkey. Invest. Ophthalmol. 11: 368-377, 1972. SCHILLER, P. H., FINLAY, B. L., AND VOLMAN, S. F. Short-term response variability of monkey striate neurons. Brain Res. 105: 347-349, 1976.
CORTEX
I
1319
17.
SCHILLER, P. H. AND KOERNER, F. Discharge characteristics of single units in superior colliculus of the alert rhesus monkey. J. Neurophysiol. 34: 920-936, 1971. 18. SCHILLER, P. H., STRYKER, M., CYNADER, M., AND BERMAN, N. Response characteristics of single cells in the monkey superior colliculus following ablation or cooling of visual cortex. J. Neurophysiol. 37: 181-194, 1974. 19. STONE, J. Morphology and physiology of the geniculocortical synapse in the cat: the question of parallel input to the striate cortex. Invest. Ophthalmol. 11: 338-346, 1972. 20. WATKINS, D. W., WILSON, J. R., SHERMAN, S. M., AND BERKLEY, M. A. Receptive field organization: further differences between simple and complex cells. Assoc. Res. Vision Ophthalmol.,
Sarasota,
21.
Fla.,
1975.
M. L., MACNICHOL, E. F., AND H. G. Glass insulated platinum microelectrode. Science 132: 1309-1310, 1960. WOLBARSHT, WAGNER,
Downloaded from http://jn.physiology.org/ at Albert R. Mann Library, Cornell University on June 3, 2013
14.
STRIATE
